diff --git a/contrast/20240905_185407.pickle b/contrast/20240905_185407.pickle new file mode 100644 index 0000000..cdae764 Binary files /dev/null and b/contrast/20240905_185407.pickle differ diff --git a/contrast/20240905_185407.txt b/contrast/20240905_185407.txt new file mode 100644 index 0000000..6311204 --- /dev/null +++ b/contrast/20240905_185407.txt @@ -0,0 +1,124 @@ +same, 6954256905319, 20240715-144627_6954256905319, 0.211, 0.750, 0.630 +same, 6930364600025, 20240719-143603_6930364600025, 0.283, 0.771, 0.616 +same, 6924509902839, 20240719-143630_6924509902839, 0.182, 0.727, 0.462 +same, 6902007115524, 20240719-143729_6902007115524, 0.221, 0.723, 0.519 +same, 6930628801472, 20240719-143807_6930628801472, 0.249, 0.804, 0.672 +same, 6908355010450, 20240719-144335_6908355010450, 0.255, 0.812, 0.575 +same, 6924509902754, 20240719-144454_6924509902754, 0.458, 0.882, 0.836 +same, 6925843410158, 20240719-144916_6925843410158, 0.229, 0.829, 0.686 +same, 6941499112529, 20240719-145022_6941499112529, 0.190, 0.795, 0.359 +same, 6918149576536, 20240719-145058_6918149576536, 0.493, 0.906, 0.765 +same, 6928002374551, 20240719-145542_6928002374551, 0.165, 0.773, 0.531 +same, 6930364600001, 20240719-150000_6930364600001, 0.322, 0.831, 0.736 +same, 6937962111526, 20240719-150209_6937962111526, 0.225, 0.794, 0.605 +same, 6924509902839, 20240719-150345_6924509902839, 0.161, 0.816, 0.502 +same, 6924509902839, 20240719-150458_6924509902839, 0.319, 0.746, 0.623 +same, 6924187895423, 20240719-150623_6924187895423, 0.430, 0.906, 0.903 +same, 6916749306102, 20240719-150716_6916749306102, 0.269, 0.764, 0.737 +same, 6951469100067, 20240719-150750_6951469100067, 0.547, 0.938, 0.896 +same, 6958350501101, 20240719-151215_6958350501101, 0.578, 0.906, 0.877 +same, 6951469100067, 20240719-151325_6951469100067, 0.610, 0.930, 0.904 +same, 6937962111526, 20240719-151349_6937962111526, 0.138, 0.762, 0.496 +same, 6916749306102, 20240719-151416_6916749306102, 0.243, 0.728, 0.630 +same, 6930364600001, 20240719-151535_6930364600001, 0.144, 0.787, 0.482 +same, 6911988027191, 20240719-151724_6911988027191, 0.035, 0.502, 0.076 +same, 6972219890632, 20240719-151800_6972219890632, 0.315, 0.880, 0.700 +same, 6901668935908, 20240719-152125_6901668935908, 0.203, 0.856, 0.426 +same, 6930364600001, 20240719-153117_6930364600001, 0.057, 0.775, 0.165 +same, 6902265587712, 20240722-144630_6902265587712, 0.364, 0.833, 0.669 +same, 4897073991658, 20240722-160713_4897073991658, 0.101, 0.514, 0.247 +same, 4892655902237, 20240722-160857_4892655902237, 0.459, 0.760, 0.812 +same, 6920907803020, 20240722-161533_6920907803020, 0.211, 0.769, 0.678 +same, 6920912342002, 20240722-163428_6920912342002, 0.281, 0.825, 0.853 +same, 6916196422325, 20240722-164859_6916196422325, 0.392, 0.837, 0.665 +same, 6901668006011, 20240722-165548_6901668006011, 0.284, 0.752, 0.530 +same, 6937221716189, 20240723-094426_6937221716189, 0.407, 0.850, 0.705 +same, 4897098800133, 20240723-094447_4897098800133, 0.382, 0.846, 0.784 +same, 6970470331529, 20240723-094617_6970470331529, 0.238, 0.925, 0.586 +same, 6918598120014, 20240723-095103_6918598120014, 0.407, 0.905, 0.819 +same, 6938737200650, 20240723-095527_6938737200650, 0.208, 0.825, 0.543 +same, 6902890022558, 20240723-095858_6902890022558, 0.162, 0.769, 0.628 +same, 6925303795535, 20240723-100031_6925303795535, 0.304, 0.835, 0.671 +same, 6920382723257, 20240723-100113_6920382723257, 0.274, 0.854, 0.539 +same, 6938737200650, 20240723-100416_6938737200650, 0.237, 0.719, 0.581 +same, 6918598120014, 20240723-100501_6918598120014, 0.099, 0.791, 0.209 +same, 6937221716189, 20240723-100726_6937221716189, 0.347, 0.771, 0.611 +same, 4897098800133, 20240723-100845_4897098800133, 0.304, 0.770, 0.718 +same, 6918598120014, 20240723-101829_6918598120014, 0.502, 0.919, 0.857 +same, 6920382723257, 20240723-102033_6920382723257, 0.343, 0.849, 0.748 +same, 6902890022558, 20240723-102206_6902890022558, 0.117, 0.745, 0.389 +same, 6970470331529, 20240723-102409_6970470331529, 0.102, 0.734, 0.305 +same, 6938737200650, 20240723-102515_6938737200650, 0.183, 0.725, 0.469 +same, 6937221716189, 20240723-102833_6937221716189, 0.067, 0.623, 0.175 +same, 6937451883132, 20240723-110216_6937451883132, 0.323, 0.860, 0.717 +same, 6973337010216, 20240723-110224_6973337010216, 0.202, 0.760, 0.622 +same, 6937451883132, 20240723-111446_6937451883132, 0.367, 0.817, 0.806 +same, 6973337010216, 20240723-111521_6973337010216, 0.140, 0.777, 0.414 +same, 6924187828544, 20240723-111602_6924187828544, 0.359, 0.859, 0.718 +same, 6934024590268, 20240723-112004_6934024590268, 0.184, 0.727, 0.583 +same, 6924187828544, 20240723-112505_6924187828544, 0.281, 0.875, 0.573 +same, 6974158894528, 20240723-155423_6974158894528, 0.433, 0.890, 0.788 +same, 6970399922518, 20240723-160227_6970399922518, 0.078, 0.564, 0.215 +same, 6974158894528, 20240723-160942_6974158894528, 0.417, 0.895, 0.832 +diff, 6908355010450, 20240715-144627_6954256905319, 0.001, 0.539, 0.003 +diff, 6924187828544, 20240719-143603_6930364600025, 0.019, 0.434, 0.038 +diff, 6918149576536, 20240719-143630_6924509902839, 0.031, 0.324, 0.062 +diff, 6918149576536, 20240719-143729_6902007115524, -0.022, 0.311, -0.040 +diff, 6924509902839, 20240719-143807_6930628801472, 0.067, 0.524, 0.144 +diff, 6937962111526, 20240719-144335_6908355010450, 0.005, 0.471, 0.015 +diff, 6901668006011, 20240719-144454_6924509902754, 0.025, 0.442, 0.046 +diff, 6970470331529, 20240719-144916_6925843410158, 0.038, 0.452, 0.096 +diff, 6911988027191, 20240719-145022_6941499112529, 0.071, 0.483, 0.153 +diff, 6930364600001, 20240719-145058_6918149576536, -0.015, 0.484, -0.034 +diff, 6930628801472, 20240719-145542_6928002374551, -0.008, 0.441, -0.026 +diff, 6924187828544, 20240719-150000_6930364600001, -0.025, 0.438, -0.052 +diff, 6925303795535, 20240719-150209_6937962111526, 0.193, 0.660, 0.414 +diff, 6924509902754, 20240719-150345_6924509902839, 0.061, 0.544, 0.182 +diff, 6928002374551, 20240719-150458_6924509902839, -0.006, 0.472, -0.016 +diff, 6902265587712, 20240719-150623_6924187895423, 0.005, 0.433, 0.010 +diff, 6924187895423, 20240719-150716_6916749306102, 0.074, 0.447, 0.167 +diff, 6941499112529, 20240719-150750_6951469100067, -0.003, 0.501, -0.007 +diff, 6902007115524, 20240719-151215_6958350501101, -0.098, 0.328, -0.193 +diff, 6924187895423, 20240719-151325_6951469100067, 0.063, 0.477, 0.112 +diff, 6941499112529, 20240719-151349_6937962111526, -0.018, 0.404, -0.047 +diff, 6937962111526, 20240719-151416_6916749306102, 0.007, 0.438, 0.023 +diff, 6918149576536, 20240719-151535_6930364600001, -0.035, 0.398, -0.082 +diff, 4897073991658, 20240719-151724_6911988027191, 0.116, 0.518, 0.219 +diff, 6901668006011, 20240719-151800_6972219890632, -0.029, 0.352, -0.062 +diff, 6902265587712, 20240719-152125_6901668935908, 0.096, 0.444, 0.184 +diff, 6918149576536, 20240719-153117_6930364600001, -0.015, 0.400, -0.031 +diff, 6954256905319, 20240722-144630_6902265587712, 0.025, 0.413, 0.072 +diff, 6928002374551, 20240722-160713_4897073991658, 0.041, 0.491, 0.131 +diff, 6924509902754, 20240722-160857_4892655902237, 0.048, 0.475, 0.094 +diff, 6951469100067, 20240722-161533_6920907803020, 0.064, 0.470, 0.129 +diff, 6937221716189, 20240722-163428_6920912342002, 0.051, 0.503, 0.122 +diff, 6920907803020, 20240722-164859_6916196422325, 0.059, 0.594, 0.157 +diff, 6972219890632, 20240722-165548_6901668006011, 0.035, 0.369, 0.067 +diff, 6916196422325, 20240723-094426_6937221716189, 0.075, 0.368, 0.107 +diff, 6930364600001, 20240723-094447_4897098800133, 0.009, 0.429, 0.019 +diff, 4892655902237, 20240723-094617_6970470331529, 0.018, 0.236, 0.037 +diff, 6973337010216, 20240723-095103_6918598120014, 0.031, 0.472, 0.084 +diff, 6928002374551, 20240723-095527_6938737200650, 0.124, 0.656, 0.348 +diff, 6974158894528, 20240723-095858_6902890022558, -0.034, 0.384, -0.090 +diff, 6908355010450, 20240723-100031_6925303795535, -0.054, 0.450, -0.104 +diff, 6901668006011, 20240723-100113_6920382723257, -0.037, 0.492, -0.072 +diff, 6908355010450, 20240723-100416_6938737200650, 0.113, 0.560, 0.222 +diff, 6937962111526, 20240723-100501_6918598120014, 0.043, 0.602, 0.126 +diff, 6920907803020, 20240723-100726_6937221716189, 0.079, 0.475, 0.181 +diff, 6925843410158, 20240723-100845_4897098800133, -0.051, 0.328, -0.144 +diff, 6924509902754, 20240723-101829_6918598120014, -0.056, 0.319, -0.092 +diff, 6970470331529, 20240723-102033_6920382723257, -0.001, 0.592, -0.003 +diff, 6954256905319, 20240723-102206_6902890022558, 0.007, 0.479, 0.021 +diff, 4897073991658, 20240723-102409_6970470331529, -0.036, 0.346, -0.101 +diff, 6951469100067, 20240723-102515_6938737200650, -0.022, 0.355, -0.036 +diff, 6924187828544, 20240723-102833_6937221716189, 0.025, 0.585, 0.068 +diff, 6954256905319, 20240723-110216_6937451883132, -0.067, 0.341, -0.181 +diff, 6930628801472, 20240723-110224_6973337010216, -0.008, 0.412, -0.023 +diff, 6951469100067, 20240723-111446_6937451883132, 0.077, 0.479, 0.119 +diff, 6958350501101, 20240723-111521_6973337010216, -0.105, 0.340, -0.184 +diff, 6925843410158, 20240723-111602_6924187828544, 0.039, 0.430, 0.096 +diff, 6928002374551, 20240723-112004_6934024590268, -0.004, 0.378, -0.010 +diff, 6916196422325, 20240723-112505_6924187828544, -0.089, 0.286, -0.140 +diff, 6973337010216, 20240723-155423_6974158894528, 0.062, 0.470, 0.139 +diff, 6925843410158, 20240723-160227_6970399922518, 0.020, 0.436, 0.056 +diff, 6924187895423, 20240723-160942_6974158894528, -0.028, 0.392, -0.048 diff --git a/contrast/20240905_185642.pickle b/contrast/20240905_185642.pickle new file mode 100644 index 0000000..a0808dc Binary files /dev/null and b/contrast/20240905_185642.pickle differ diff --git a/contrast/20240905_185642.txt b/contrast/20240905_185642.txt new file mode 100644 index 0000000..dda7cbf --- /dev/null +++ b/contrast/20240905_185642.txt @@ -0,0 +1,124 @@ +same, 6954256905319, 20240715-144627_6954256905319, 0.211, 0.750, 0.630 +same, 6930364600025, 20240719-143603_6930364600025, 0.283, 0.771, 0.616 +same, 6924509902839, 20240719-143630_6924509902839, 0.182, 0.727, 0.462 +same, 6902007115524, 20240719-143729_6902007115524, 0.221, 0.723, 0.519 +same, 6930628801472, 20240719-143807_6930628801472, 0.249, 0.804, 0.672 +same, 6908355010450, 20240719-144335_6908355010450, 0.255, 0.812, 0.575 +same, 6924509902754, 20240719-144454_6924509902754, 0.458, 0.882, 0.836 +same, 6925843410158, 20240719-144916_6925843410158, 0.229, 0.829, 0.686 +same, 6941499112529, 20240719-145022_6941499112529, 0.190, 0.795, 0.359 +same, 6918149576536, 20240719-145058_6918149576536, 0.493, 0.906, 0.765 +same, 6928002374551, 20240719-145542_6928002374551, 0.165, 0.773, 0.531 +same, 6930364600001, 20240719-150000_6930364600001, 0.322, 0.831, 0.736 +same, 6937962111526, 20240719-150209_6937962111526, 0.225, 0.794, 0.605 +same, 6924509902839, 20240719-150345_6924509902839, 0.161, 0.816, 0.502 +same, 6924509902839, 20240719-150458_6924509902839, 0.319, 0.746, 0.623 +same, 6924187895423, 20240719-150623_6924187895423, 0.430, 0.906, 0.903 +same, 6916749306102, 20240719-150716_6916749306102, 0.269, 0.764, 0.737 +same, 6951469100067, 20240719-150750_6951469100067, 0.547, 0.938, 0.896 +same, 6958350501101, 20240719-151215_6958350501101, 0.578, 0.906, 0.877 +same, 6951469100067, 20240719-151325_6951469100067, 0.610, 0.930, 0.904 +same, 6937962111526, 20240719-151349_6937962111526, 0.138, 0.762, 0.496 +same, 6916749306102, 20240719-151416_6916749306102, 0.243, 0.728, 0.630 +same, 6930364600001, 20240719-151535_6930364600001, 0.144, 0.787, 0.482 +same, 6911988027191, 20240719-151724_6911988027191, 0.035, 0.502, 0.076 +same, 6972219890632, 20240719-151800_6972219890632, 0.315, 0.880, 0.700 +same, 6901668935908, 20240719-152125_6901668935908, 0.203, 0.856, 0.426 +same, 6930364600001, 20240719-153117_6930364600001, 0.057, 0.775, 0.165 +same, 6902265587712, 20240722-144630_6902265587712, 0.364, 0.833, 0.669 +same, 4897073991658, 20240722-160713_4897073991658, 0.101, 0.514, 0.247 +same, 4892655902237, 20240722-160857_4892655902237, 0.459, 0.760, 0.812 +same, 6920907803020, 20240722-161533_6920907803020, 0.211, 0.769, 0.678 +same, 6920912342002, 20240722-163428_6920912342002, 0.281, 0.825, 0.853 +same, 6916196422325, 20240722-164859_6916196422325, 0.392, 0.837, 0.665 +same, 6901668006011, 20240722-165548_6901668006011, 0.284, 0.752, 0.530 +same, 6937221716189, 20240723-094426_6937221716189, 0.407, 0.850, 0.705 +same, 4897098800133, 20240723-094447_4897098800133, 0.382, 0.846, 0.784 +same, 6970470331529, 20240723-094617_6970470331529, 0.238, 0.925, 0.586 +same, 6918598120014, 20240723-095103_6918598120014, 0.407, 0.905, 0.819 +same, 6938737200650, 20240723-095527_6938737200650, 0.208, 0.825, 0.543 +same, 6902890022558, 20240723-095858_6902890022558, 0.162, 0.769, 0.628 +same, 6925303795535, 20240723-100031_6925303795535, 0.304, 0.835, 0.671 +same, 6920382723257, 20240723-100113_6920382723257, 0.274, 0.854, 0.539 +same, 6938737200650, 20240723-100416_6938737200650, 0.237, 0.719, 0.581 +same, 6918598120014, 20240723-100501_6918598120014, 0.099, 0.791, 0.209 +same, 6937221716189, 20240723-100726_6937221716189, 0.347, 0.771, 0.611 +same, 4897098800133, 20240723-100845_4897098800133, 0.304, 0.770, 0.718 +same, 6918598120014, 20240723-101829_6918598120014, 0.502, 0.919, 0.857 +same, 6920382723257, 20240723-102033_6920382723257, 0.343, 0.849, 0.748 +same, 6902890022558, 20240723-102206_6902890022558, 0.117, 0.745, 0.389 +same, 6970470331529, 20240723-102409_6970470331529, 0.102, 0.734, 0.305 +same, 6938737200650, 20240723-102515_6938737200650, 0.183, 0.725, 0.469 +same, 6937221716189, 20240723-102833_6937221716189, 0.067, 0.623, 0.175 +same, 6937451883132, 20240723-110216_6937451883132, 0.323, 0.860, 0.717 +same, 6973337010216, 20240723-110224_6973337010216, 0.202, 0.760, 0.622 +same, 6937451883132, 20240723-111446_6937451883132, 0.367, 0.817, 0.806 +same, 6973337010216, 20240723-111521_6973337010216, 0.140, 0.777, 0.414 +same, 6924187828544, 20240723-111602_6924187828544, 0.359, 0.859, 0.718 +same, 6934024590268, 20240723-112004_6934024590268, 0.184, 0.727, 0.583 +same, 6924187828544, 20240723-112505_6924187828544, 0.281, 0.875, 0.573 +same, 6974158894528, 20240723-155423_6974158894528, 0.433, 0.890, 0.788 +same, 6970399922518, 20240723-160227_6970399922518, 0.078, 0.564, 0.215 +same, 6974158894528, 20240723-160942_6974158894528, 0.417, 0.895, 0.832 +diff, 6972219890632, 20240715-144627_6954256905319, 0.045, 0.441, 0.100 +diff, 4897073991658, 20240719-143603_6930364600025, 0.030, 0.305, 0.060 +diff, 6934024590268, 20240719-143630_6924509902839, -0.003, 0.922, -0.011 +diff, 6920382723257, 20240719-143729_6902007115524, 0.064, 0.471, 0.140 +diff, 6902890022558, 20240719-143807_6930628801472, 0.025, 0.560, 0.083 +diff, 6930364600001, 20240719-144335_6908355010450, 0.024, 0.445, 0.063 +diff, 6924509902839, 20240719-144454_6924509902754, 0.147, 0.521, 0.281 +diff, 6918598120014, 20240719-144916_6925843410158, -0.119, 0.340, -0.288 +diff, 6920382723257, 20240719-145022_6941499112529, 0.099, 0.417, 0.185 +diff, 6902890022558, 20240719-145058_6918149576536, 0.006, 0.453, 0.019 +diff, 6902007115524, 20240719-145542_6928002374551, -0.011, 0.392, -0.028 +diff, 6958350501101, 20240719-150000_6930364600001, -0.030, 0.354, -0.049 +diff, 6937451883132, 20240719-150209_6937962111526, -0.039, 0.302, -0.085 +diff, 6916749306102, 20240719-150345_6924509902839, 0.125, 0.664, 0.441 +diff, 6972219890632, 20240719-150458_6924509902839, 0.143, 0.548, 0.272 +diff, 6924509902754, 20240719-150623_6924187895423, 0.016, 0.506, 0.034 +diff, 6958350501101, 20240719-150716_6916749306102, -0.015, 0.388, -0.030 +diff, 6973337010216, 20240719-150750_6951469100067, 0.013, 0.473, 0.037 +diff, 4897073991658, 20240719-151215_6958350501101, 0.057, 0.383, 0.107 +diff, 4897098800133, 20240719-151325_6951469100067, 0.164, 0.467, 0.321 +diff, 6920912342002, 20240719-151349_6937962111526, -0.005, 0.477, -0.016 +diff, 4897098800133, 20240719-151416_6916749306102, -0.052, 0.289, -0.122 +diff, 6973337010216, 20240719-151535_6930364600001, 0.036, 0.507, 0.146 +diff, 6908355010450, 20240719-151724_6911988027191, -0.031, 0.363, -0.057 +diff, 6924187895423, 20240719-151800_6972219890632, 0.056, 0.421, 0.117 +diff, 6928002374551, 20240719-152125_6901668935908, -0.007, 0.405, -0.020 +diff, 6954256905319, 20240719-153117_6930364600001, -0.016, 0.470, -0.057 +diff, 6902890022558, 20240722-144630_6902265587712, -0.000, 0.477, -0.001 +diff, 6918149576536, 20240722-160713_4897073991658, -0.043, 0.421, -0.083 +diff, 6924509902839, 20240722-160857_4892655902237, 0.082, 0.573, 0.167 +diff, 6920912342002, 20240722-161533_6920907803020, 0.016, 0.488, 0.049 +diff, 6916749306102, 20240722-163428_6920912342002, 0.072, 0.584, 0.206 +diff, 6901668935908, 20240722-164859_6916196422325, -0.060, 0.385, -0.124 +diff, 6970399922518, 20240722-165548_6901668006011, 0.024, 0.378, 0.055 +diff, 6924509902754, 20240723-094426_6937221716189, -0.109, 0.329, -0.190 +diff, 6938737200650, 20240723-094447_4897098800133, 0.039, 0.540, 0.093 +diff, 6901668006011, 20240723-094617_6970470331529, -0.074, 0.426, -0.168 +diff, 4897073991658, 20240723-095103_6918598120014, 0.007, 0.402, 0.014 +diff, 6925843410158, 20240723-095527_6938737200650, -0.035, 0.402, -0.096 +diff, 6901668935908, 20240723-095858_6902890022558, -0.062, 0.342, -0.145 +diff, 6951469100067, 20240723-100031_6925303795535, 0.002, 0.396, 0.003 +diff, 4897098800133, 20240723-100113_6920382723257, 0.101, 0.407, 0.211 +diff, 6925843410158, 20240723-100416_6938737200650, -0.068, 0.347, -0.172 +diff, 6908355010450, 20240723-100501_6918598120014, -0.077, 0.424, -0.158 +diff, 6920907803020, 20240723-100726_6937221716189, 0.079, 0.475, 0.181 +diff, 6973337010216, 20240723-100845_4897098800133, 0.023, 0.449, 0.070 +diff, 6920907803020, 20240723-101829_6918598120014, 0.140, 0.673, 0.296 +diff, 4897073991658, 20240723-102033_6920382723257, 0.019, 0.319, 0.043 +diff, 6908355010450, 20240723-102206_6902890022558, 0.060, 0.429, 0.123 +diff, 4892655902237, 20240723-102409_6970470331529, 0.013, 0.320, 0.033 +diff, 6925843410158, 20240723-102515_6938737200650, -0.005, 0.406, -0.013 +diff, 6916749306102, 20240723-102833_6937221716189, 0.137, 0.625, 0.424 +diff, 6930364600001, 20240723-110216_6937451883132, -0.018, 0.387, -0.040 +diff, 6911988027191, 20240723-110224_6973337010216, -0.017, 0.518, -0.045 +diff, 6925843410158, 20240723-111446_6937451883132, -0.026, 0.382, -0.066 +diff, 6911988027191, 20240723-111521_6973337010216, -0.003, 0.430, -0.008 +diff, 6924509902754, 20240723-111602_6924187828544, -0.058, 0.322, -0.109 +diff, 6902265587712, 20240723-112004_6934024590268, -0.020, 0.345, -0.039 +diff, 6901668935908, 20240723-112505_6924187828544, 0.039, 0.415, 0.074 +diff, 6908355010450, 20240723-155423_6974158894528, 0.018, 0.420, 0.029 +diff, 6937962111526, 20240723-160227_6970399922518, -0.026, 0.378, -0.084 +diff, 6951469100067, 20240723-160942_6974158894528, 0.035, 0.375, 0.050 diff --git a/contrast/__pycache__/config.cpython-39.pyc b/contrast/__pycache__/config.cpython-39.pyc index ba23f1d..e81b06c 100644 Binary files a/contrast/__pycache__/config.cpython-39.pyc and b/contrast/__pycache__/config.cpython-39.pyc differ diff --git a/contrast/__pycache__/test_ori.cpython-39.pyc b/contrast/__pycache__/test_ori.cpython-39.pyc new file mode 100644 index 0000000..f2ab6ae Binary files /dev/null and b/contrast/__pycache__/test_ori.cpython-39.pyc differ diff --git a/contrast/checkpoints/resnet18_0515/best.rknn b/contrast/checkpoints/resnet18_0515/best.rknn new file mode 100644 index 0000000..62fe8fe Binary files /dev/null and b/contrast/checkpoints/resnet18_0515/best.rknn differ diff --git a/contrast/config.py b/contrast/config.py index aea7680..4864cdb 100644 --- a/contrast/config.py +++ b/contrast/config.py @@ -4,15 +4,13 @@ import torchvision.transforms as T class Config: # network settings - backbone = 'vit' # [resnet18, mobilevit_s, mobilenet_v2, mobilenetv3_small, mobilenetv3_large, mobilenet_v1, PPLCNET_x1_0, PPLCNET_x0_5, PPLCNET_x2_5] - metric = 'softmax' # [cosface, arcface, softmax] + backbone = 'resnet18' # [resnet18, mobilevit_s, mobilenet_v2, mobilenetv3_small, mobilenetv3_large, mobilenet_v1, PPLCNET_x1_0, PPLCNET_x0_5, PPLCNET_x2_5] + metric = 'arcface' # [cosface, arcface] cbam = True - embedding_size = 256 # 256 + embedding_size = 256 drop_ratio = 0.5 img_size = 224 - teacher = 'vit' # [resnet18, mobilevit_s, mobilenet_v2, mobilenetv3_small, mobilenetv3_large, mobilenet_v1, PPLCNET_x1_0, PPLCNET_x0_5, PPLCNET_x2_5] - student = 'resnet' # data preprocess # input_shape = [1, 128, 128] """transforms.RandomCrop(size), @@ -24,7 +22,7 @@ class Config: train_transform = T.Compose([ T.ToTensor(), T.Resize((img_size, img_size)), - # T.RandomCrop(img_size*4//5), + # T.RandomCrop(img_size), # T.RandomHorizontalFlip(p=0.5), T.RandomRotation(180), T.ColorJitter(brightness=0.5), @@ -39,39 +37,39 @@ class Config: ]) # dataset - train_root = './data/2250_train/train' # 初始筛选过一次的数据集 - # train_root = './data/0625_train/train' + train_root = './data/2250_train/train' # 初始筛选过一次的数据集 + # train_root = './data/0612_train/train' test_root = "./data/2250_train/val/" - # test_root = "./data/0625_train/val" - + # test_root = "./data/0612_train/val" test_list = "./data/2250_train/val_pair.txt" - test_group_json = "./data/2250_train/cross_same.json" - # test_group_json = "./data/0625_train/cross_same.json" + + test_group_json = "./2250_train/cross_same_0508.json" + + # test_list = "./data/test_data_100/val_pair.txt" # training settings - checkpoints = "checkpoints/vit_b_16_0815/" # [resnet18, mobilevit_s, mobilenet_v2, mobilenetv3] - restore = True + checkpoints = "checkpoints/resnet18_0613/" # [resnet18, mobilevit_s, mobilenet_v2, mobilenetv3] + restore = False # restore_model = "checkpoints/renet18_2250_0315/best_resnet18_2250_0315.pth" # best_resnet18_1491_0306.pth - restore_model = "checkpoints/vit_b_16_0730/best.pth" # best_resnet18_1491_0306.pth + restore_model = "checkpoints/resnet18_0515/best.pth" # best_resnet18_1491_0306.pth - # test_model = "./checkpoints/renet18_1887_0311/best_resnet18_1887_0311.pth" + # test_model = "checkpoints/renet18_2250_0314/best_resnet18_2250_0314.pth" testbackbone = 'resnet18' # [resnet18, mobilevit_s, mobilenet_v2, mobilenetv3_small, mobilenetv3_large, mobilenet_v1, PPLCNET_x1_0, PPLCNET_x0_5] - # test_val = "./data/2250_train" - test_val = "./data/0625_train" - test_model = "checkpoints/resnet18_0721/best.pth" + test_val = "D:/比对/cl" + # test_val = "./data/test_data_100" + test_model = "checkpoints/resnet18_0515/best.pth" - train_batch_size = 128 # 256 + train_batch_size = 512 # 256 test_batch_size = 256 # 256 - epoch = 300 - optimizer = 'adamw' # ['sgd', 'adam'， 'adamw'] - lr = 1e-3 # 1e-2 - lr_step = 10 # 10 + optimizer = 'sgd' # ['sgd', 'adam'] + lr = 1.5e-2 # 1e-2 + lr_step = 5 # 10 lr_decay = 0.95 # 0.98 weight_decay = 5e-4 - loss = 'focal_loss' # ['focal_loss', 'cross_entropy'] + loss = 'cross_entropy' # ['focal_loss', 'cross_entropy'] device = torch.device('cuda:1' if torch.cuda.is_available() else 'cpu') # device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu') @@ -79,6 +77,5 @@ class Config: num_workers = 4 # dataloader group_test = True - # group_test = False -config = Config() \ No newline at end of file +config = Config() diff --git a/contrast/contrast_one2one.py b/contrast/contrast_one2one.py index 040865e..f3e9ad5 100644 --- a/contrast/contrast_one2one.py +++ b/contrast/contrast_one2one.py @@ -33,150 +33,176 @@ import numpy as np import cv2 import os import sys - +import random import pickle import torch import time import json -from config import config as conf -from model import resnet18 -from inference import load_contrast_model -from inference import featurize +from pathlib import Path +from scipy.spatial.distance import cdist +import matplotlib.pyplot as plt +import shutil +from datetime import datetime +# Vit版resnet, 和现场特征不一致，需将resnet_vit中文件提出 +# from config import config as conf +# from model import resnet18 +# from inference import load_contrast_model +# from inference import featurize +# embedding_size = conf.embedding_size +# img_size = conf.img_size +# device = conf.device +# model = load_contrast_model() sys.path.append(r"D:\DetectTracking") from tracking.utils.read_data import extract_data, read_tracking_output, read_deletedBarcode_file +from config import config as conf +from model import resnet18 as resnet18 +from test_ori import inference_image + + IMG_FORMAT = ['.bmp', '.jpg', '.jpeg', '.png'] +model = resnet18().to(conf.device) -model = load_contrast_model() +print('load model {} '.format(conf.testbackbone)) +# model = nn.DataParallel(model).to(conf.device) +model.load_state_dict(torch.load(conf.test_model, map_location=conf.device)) +model.eval() -def creat_shopping_event(basePath, savePath, subimgPath=False): - eventList = [] + +def creat_shopping_event(eventPath, subimgPath=False): + '''构造放入商品事件字典，这些事件需满足条件： + 1) 前后摄至少有一条轨迹输出 + 2) 保存有帧图像，以便裁剪出 boxe 子图 + ''' + # filename = "20240723-155413_6904406215720" - '''一、构造放入商品事件列表''' - k = 0 - for filename in os.listdir(basePath): - # filename = "20240723-155413_6904406215720" - - '''filename下为一次购物事件''' - filepath = os.path.join(basePath, filename) + '''filename下为一次购物事件''' + eventName = os.path.basename(eventPath) - '''================ 0. 检查 filename 及 filepath 正确性和有效性 ================''' - nmlist = filename.split('_') - if filename.find('2024')<0 or len(nmlist)!=2 or len(nmlist[0])!=15 or len(nmlist[1])<11: - continue - if not os.path.isdir(filepath): continue + '''================ 0. 检查 filename 及 eventPath 正确性和有效性 ================''' + nmlist = eventName.split('_') + if eventName.find('2024')<0 or len(nmlist)!=2 or len(nmlist[0])!=15 or len(nmlist[1])<11: + return + if not os.path.isdir(eventPath): + return + + + + '''================ 1. 构造事件描述字典，暂定 9 items ===============''' + event = {} + event['barcode'] = eventName.split('_')[1] + event['type'] = 'input' + event['filepath'] = eventPath + event['back_imgpaths'] = [] + event['front_imgpaths'] = [] + event['back_boxes'] = np.empty((0, 9), dtype=np.float64) + event['front_boxes'] = np.empty((0, 9), dtype=np.float64) + event['back_feats'] = np.empty((0, 256), dtype=np.float64) + event['front_feats'] = np.empty((0, 256), dtype=np.float64) + event['feats_compose'] = np.empty((0, 256), dtype=np.float64) + # event['feats_select'] = np.empty((0, 256), dtype=np.float64) + + + '''================= 2. 读取 data 文件 =============================''' + for dataname in os.listdir(eventPath): + # filename = '1_track.data' + datapath = os.path.join(eventPath, dataname) + if not os.path.isfile(datapath): continue - '''================ 1. 构造事件描述字典，暂定 9 items ===============''' - event = {} - event['barcode'] = nmlist[1] - event['type'] = 'input' - event['filepath'] = filepath - event['back_imgpaths'] = [] - event['front_imgpaths'] = [] - event['back_boxes'] = np.empty((0, 9), dtype=np.float64) - event['front_boxes'] = np.empty((0, 9), dtype=np.float64) - event['back_feats'] = np.empty((0, 256), dtype=np.float64) - event['front_feats'] = np.empty((0, 256), dtype=np.float64) - event['feats_compose'] = np.empty((0, 256), dtype=np.float64) - # event['feats_select'] = np.empty((0, 256), dtype=np.float64) - - - '''================= 2. 读取 data 文件 =============================''' - for dataname in os.listdir(filepath): - # filename = '1_track.data' - datapath = os.path.join(filepath, dataname) - if not os.path.isfile(datapath): continue - - CamerType = dataname.split('_')[0] - ''' 2.1 读取 0/1_track.data 中数据，暂不考虑''' - # if dataname.find("_track.data")>0: - # bboxes, ffeats, trackerboxes, tracker_feat_dict, trackingboxes, tracking_feat_dict = extract_data(datapath) + CamerType = dataname.split('_')[0] + ''' 2.1 读取 0/1_track.data 中数据，暂不考虑''' + # if dataname.find("_track.data")>0: + # bboxes, ffeats, trackerboxes, tracker_feat_dict, trackingboxes, tracking_feat_dict = extract_data(datapath) - ''' 2.2 读取 0/1_tracking_output.data 中数据''' - if dataname.find("_tracking_output.data")>0: - tracking_output_boxes, tracking_output_feats = read_tracking_output(datapath) - if len(tracking_output_boxes) != len(tracking_output_feats): continue - if CamerType == '0': - event['back_boxes'] = tracking_output_boxes - event['back_feats'] = tracking_output_feats - elif CamerType == '1': - event['front_boxes'] = tracking_output_boxes - event['front_feats'] = tracking_output_feats - - '''2.3 事件的特征表征方式: 特征选择、特征集成''' - bk_feats = event['back_feats'] - ft_feats = event['front_feats'] - - '''2.3.1 特征集成''' - feats_compose = np.empty((0, 256), dtype=np.float64) - if len(ft_feats): - feats_compose = np.concatenate((feats_compose, ft_feats), axis=0) - if len(bk_feats): - feats_compose = np.concatenate((feats_compose, bk_feats), axis=0) - event['feats_compose'] = feats_compose - - '''2.3.1 特征选择''' - if len(ft_feats): - event['feats_select'] = ft_feats - - pickpath = os.path.join(savePath, f"{filename}.pickle") - with open(pickpath, 'wb') as f: - pickle.dump(event, f) - print(f"Event: {filename}") - - if subimgPath==False: - eventList.append(event) - continue - - '''================ 2. 读取图像文件地址，并按照帧ID排序 =============''' - frontImgs, frontFid = [], [] - backImgs, backFid = [], [] - for imgname in os.listdir(filepath): - name, ext = os.path.splitext(imgname) - if ext not in IMG_FORMAT or name.find('frameId')<0: continue - - CamerType = name.split('_')[0] - frameId = int(name.split('_')[3]) - imgpath = os.path.join(filepath, imgname) + ''' 2.2 读取 0/1_tracking_output.data 中数据''' + if dataname.find("_tracking_output.data")>0: + tracking_output_boxes, tracking_output_feats = read_tracking_output(datapath) + if len(tracking_output_boxes) != len(tracking_output_feats): continue if CamerType == '0': - backImgs.append(imgpath) - backFid.append(frameId) - if CamerType == '1': - frontImgs.append(imgpath) - frontFid.append(frameId) - - frontIdx = np.argsort(np.array(frontFid)) - backIdx = np.argsort(np.array(backFid)) - - '''2.1 生成依据帧 ID 排序的前后摄图像地址列表''' - frontImgs = [frontImgs[i] for i in frontIdx] - backImgs = [backImgs[i] for i in backIdx] + event['back_boxes'] = tracking_output_boxes + event['back_feats'] = tracking_output_feats + elif CamerType == '1': + event['front_boxes'] = tracking_output_boxes + event['front_feats'] = tracking_output_feats + + + if len(event['back_boxes'])==0 or len(event['front_boxes'])==0: + return None + + '''2.3 事件的特征表征方式: 特征选择、特征集成''' + bk_feats = event['back_feats'] + ft_feats = event['front_feats'] + + '''2.3.1 特征集成''' + feats_compose = np.empty((0, 256), dtype=np.float64) + if len(ft_feats): + feats_compose = np.concatenate((feats_compose, ft_feats), axis=0) + if len(bk_feats): + feats_compose = np.concatenate((feats_compose, bk_feats), axis=0) + event['feats_compose'] = feats_compose + + '''2.3.1 特征选择''' + if len(ft_feats): + event['feats_select'] = ft_feats - '''2.2 将前、后摄图像路径添加至事件字典''' - bfid = event['back_boxes'][:, 7].astype(np.int64) - ffid = event['front_boxes'][:, 7].astype(np.int64) - if len(bfid) and max(bfid) <= len(backImgs): - event['back_imgpaths'] = [backImgs[i-1] for i in bfid] - if len(ffid) and max(ffid) <= len(frontImgs): - event['front_imgpaths'] = [frontImgs[i-1] for i in ffid] + # pickpath = os.path.join(savePath, f"{filename}.pickle") + # with open(pickpath, 'wb') as f: + # pickle.dump(event, f) + # print(f"Event: {filename}") + + # if subimgPath==False: + # eventList.append(event) + # continue - - '''================ 3. 判断当前事件有效性，并添加至事件列表 ==========''' - condt1 = len(event['back_imgpaths'])==0 or len(event['front_imgpaths'])==0 - condt2 = len(event['front_feats'])==0 and len(event['back_feats'])==0 - if condt1 or condt2: - print(f" Error, condt1: {condt1}, condt2: {condt2}") - continue + '''================ 2. 读取图像文件地址，并按照帧ID排序 =============''' + frontImgs, frontFid = [], [] + backImgs, backFid = [], [] + for imgname in os.listdir(eventPath): + name, ext = os.path.splitext(imgname) + if ext not in IMG_FORMAT or name.find('frameId')<0: continue - eventList.append(event) - - # k += 1 - # if k==1: - # continue - - '''一、构造放入商品事件列表，暂不处理''' + CamerType = name.split('_')[0] + frameId = int(name.split('_')[3]) + imgpath = os.path.join(eventPath, imgname) + if CamerType == '0': + backImgs.append(imgpath) + backFid.append(frameId) + if CamerType == '1': + frontImgs.append(imgpath) + frontFid.append(frameId) + + frontIdx = np.argsort(np.array(frontFid)) + backIdx = np.argsort(np.array(backFid)) + + '''2.1 生成依据帧 ID 排序的前后摄图像地址列表''' + frontImgs = [frontImgs[i] for i in frontIdx] + backImgs = [backImgs[i] for i in backIdx] + + '''2.2 将前、后摄图像路径添加至事件字典''' + + + bfid = event['back_boxes'][:, 7].astype(np.int64) + ffid = event['front_boxes'][:, 7].astype(np.int64) + if len(bfid) and max(bfid) <= len(backImgs): + event['back_imgpaths'] = [backImgs[i-1] for i in bfid] + if len(ffid) and max(ffid) <= len(frontImgs): + event['front_imgpaths'] = [frontImgs[i-1] for i in ffid] + + + '''================ 3. 判断当前事件有效性，并添加至事件列表 ==========''' + condt1 = len(event['back_imgpaths'])==0 or len(event['front_imgpaths'])==0 + condt2 = len(event['front_feats'])==0 and len(event['back_feats'])==0 + + if condt1 or condt2: + print(f"Event: {eventName}, Error, condt1: {condt1}, condt2: {condt2}") + return None + + + + + '''构造放入商品事件列表，暂不处理''' # delepath = os.path.join(basePath, 'deletedBarcode.txt') # bcdList = read_deletedBarcode_file(delepath) # for slist in bcdList: @@ -192,9 +218,8 @@ def creat_shopping_event(basePath, savePath, subimgPath=False): # event['barcode'] = slist['Deleted'].strip() # event['type'] = 'getout' # event['basePath'] = getoutPath - - return eventList + return event def get_std_barcodeDict(bcdpath, bpath): ''' @@ -206,7 +231,7 @@ def get_std_barcodeDict(bcdpath, bpath): bpath: 字典存储地址 ''' - # bpath = r'\\192.168.1.28\share\测试_202406\contrast\barcodes' + # bpath = r'\\192.168.1.28\share\测试_202406\contrast\std_barcodes' '''读取数据集中 barcode 列表''' stdBlist = [] @@ -255,61 +280,41 @@ def get_std_barcodeDict(bcdpath, bpath): return - -def extract_save_trajture_subimgs(shoppingEventPath, shoppingFeatPath, subimgPath): - '''用于保存一次购物事件的轨迹图像子图''' - - shoppingFeatPath = r"\\192.168.1.28\share\测试_202406\contrast\events" - subimgPath = r'\\192.168.1.28\share\测试_202406\contrast\subimgs' - - - eventList = creat_shopping_event(shoppingEventPath, shoppingFeatPath, subimgPath=True) - print("======= eventList have generated and features have saved! =======") - - - barcodeDict = {} - for event in eventList: - '''9 items: barcode, type, filepath, back_imgpaths, front_imgpaths, - back_boxes, front_boxes, back_feats, front_feats - ''' - if len(event['feats_select']): - event_feats = event['feats_select'] - elif len(event['back_feats']): - event_feats = event['back_feats'] +def save_event_subimg(event, savepath): + ''' + 功能: 保存一次购物事件的轨迹子图 + 9 items: barcode, type, filepath, back_imgpaths, front_imgpaths, + back_boxes, front_boxes, back_feats, front_feats, + feats_compose, feats_select + 子图保存次序：先前摄、后后摄，以 k 为编号，和 "feats_compose" 中次序相同 + ''' + cameras = ('front', 'back') + k = 0 + for camera in cameras: + if camera == 'front': + boxes = event['front_boxes'] + imgpaths = event['front_imgpaths'] else: - continue + boxes = event['back_boxes'] + imgpaths = event['back_imgpaths'] - - '''保存一次购物事件的轨迹子图''' - basename = os.path.basename(event['filepath']) - spath = os.path.join(subimgPath, basename) - if not os.path.exists(spath): - os.makedirs(spath) - cameras = ('front', 'back') - for camera in cameras: - if camera == 'front': - boxes = event['front_boxes'] - imgpaths = event['front_imgpaths'] - else: - boxes = event['back_boxes'] - imgpaths = event['back_imgpaths'] - - for i, box in enumerate(boxes): - x1, y1, x2, y2, tid, score, cls, fid, bid = box - - imgpath = imgpaths[i] - image = cv2.imread(imgpath) - subimg = image[int(y1/2):int(y2/2), int(x1/2):int(x2/2), :] - - camerType, timeTamp, _, frameID = os.path.basename(imgpath).split('.')[0].split('_') - subimgName = f"{camerType}_{tid}_fid({fid}, {frameID}).png" - subimgPath = os.path.join(spath, subimgName) - - cv2.imwrite(subimgPath, subimg) - - print(f"Image saved: {basename}") + for i, box in enumerate(boxes): + x1, y1, x2, y2, tid, score, cls, fid, bid = box - + imgpath = imgpaths[i] + image = cv2.imread(imgpath) + subimg = image[int(y1/2):int(y2/2), int(x1/2):int(x2/2), :] + + camerType, timeTamp, _, frameID = os.path.basename(imgpath).split('.')[0].split('_') + subimgName = f"{k}_cam-{camerType}_tid-{int(tid)}_fid-({int(fid)}, {frameID}).png" + spath = os.path.join(savepath, subimgName) + + cv2.imwrite(spath, subimg) + k += 1 + # basename = os.path.basename(event['filepath']) + print(f"Image saved: {os.path.basename(event['filepath'])}") + + def batch_inference(imgpaths, batch): size = len(imgpaths) @@ -326,7 +331,7 @@ def batch_inference(imgpaths, batch): return features -def stdfeat_infer(imgPath, featPath): +def stdfeat_infer(imgPath, featPath, bcdSet=None): ''' inputs: imgPath: 该文件夹下的 pickle 文件格式 {barcode: [imgpath1, imgpath1, ...]} @@ -337,11 +342,15 @@ def stdfeat_infer(imgPath, featPath): ''' - # imgPath = r"\\192.168.1.28\share\测试_202406\contrast\barcodes" - # featPath = r"\\192.168.1.28\share\测试_202406\contrast\features" + # imgPath = r"\\192.168.1.28\share\测试_202406\contrast\std_barcodes" + # featPath = r"\\192.168.1.28\share\测试_202406\contrast\std_features" stdBarcodeDict = {} k = 0 for filename in os.listdir(imgPath): + bcd, ext = os.path.splitext(filename) + if bcdSet is not None and bcd not in bcdSet: + continue + filepath = os.path.join(imgPath, filename) stdbDict = {} @@ -351,12 +360,21 @@ def stdfeat_infer(imgPath, featPath): with open(filepath, 'rb') as f: bpDict = pickle.load(f) for barcode, imgpaths in bpDict.items(): - feature = batch_inference(imgpaths, 8) + # feature = batch_inference(imgpaths, 8) #from vit distilled model of Lichen + feature = inference_image(imgpaths, conf.test_transform, model, conf.device) + + feature /= np.linalg.norm(feature, axis=1)[:, None] + except Exception as e: print(f"Error accured at: {filename}, with Exception is: {e}") '''================ 保存单个barcode特征 ================''' - stdbDict[barcode] = feature + stdbDict["barcode"] = barcode + stdbDict["imgpaths"] = imgpaths + stdbDict["feats"] = feature + + + pkpath = os.path.join(featPath, f"{barcode}.pickle") with open(pkpath, 'wb') as f: pickle.dump(stdbDict, f) @@ -364,25 +382,210 @@ def stdfeat_infer(imgPath, featPath): stdBarcodeDict[barcode] = feature t2 = time.time() print(f"Barcode: {barcode}, need time: {t2-t1:.1f} secs") - k += 1 - if k == 10: - break + # k += 1 + # if k == 10: + # break pickpath = os.path.join(featPath, f"barcode_features_{k}.pickle") with open(pickpath, 'wb') as f: pickle.dump(stdBarcodeDict, f) -def contrast_performance_evaluate(): - eventFeatPath = r"\\192.168.1.28\share\测试_202406\contrast\events" - stdFeatPath = r"\\192.168.1.28\share\测试_202406\contrast\features" - + -def generate_event_and_standard_features(): + +def contrast_performance_evaluate(): + eventFeatPath = r"\\192.168.1.28\share\测试_202406\contrast\events" + stdBcdPath = r"\\192.168.1.28\share\测试_202406\contrast\std_barcodes" + + stdFeatPath = r"\\192.168.1.28\share\测试_202406\contrast\std_features" + + subimgPath = r"\\192.168.1.28\share\测试_202406\contrast\subimgs" + + + # stdBarcode = [p.stem for p in Path(stdFeatPath).iterdir() if p.is_file() and p.suffix=='.pickle'] + stdBarcode = [p.stem for p in Path(stdBcdPath).iterdir() if p.is_file() and p.suffix=='.pickle'] + + + '''购物事件列表，该列表中的 Barcode 存在于标准的 stdBarcode 内''' + # evtList = [(p.stem, p.stem.split('_')[1]) for p in Path(eventFeatPath).iterdir() + # if p.is_file() + # and p.suffix=='.pickle' + # and len(p.stem.split('_'))==2 + # and p.stem.split('_')[1].isdigit() + # and p.stem.split('_')[1] in stdBarcode + # ] + + evtList = [(p.stem, p.stem.split('_')[1]) for p in Path(eventFeatPath).iterdir() + if p.is_file() + and p.suffix=='.pickle' + and len(p.stem.split('_'))==2 + and p.stem.split('_')[1].isdigit() + and p.stem.split('_')[1] in stdBarcode + ] + + barcodes = set([bcd for _, bcd in evtList]) + + '''标准特征集图像样本经特征提取并保存，运行一次后无需再运行''' + # stdfeat_infer(stdBcdPath, stdFeatPath, barcodes) + + '''========= 构建用于比对的标准特征字典 =============''' + stdDict = {} + for barcode in barcodes: + stdpath = os.path.join(stdFeatPath, barcode+'.pickle') + with open(stdpath, 'rb') as f: + stddata = pickle.load(f) + stdDict[barcode] = stddata + + '''========= 构建用于比对的操作事件字典 =============''' + evtDict = {} + for event, barcode in evtList: + evtpath = os.path.join(eventFeatPath, event+'.pickle') + with open(evtpath, 'rb') as f: + evtdata = pickle.load(f) + evtDict[event] = evtdata + + + '''===== 构造 3 个事件对: 扫 A 放 A, 扫 A 放 B, 合并 ====================''' + AA_list = [(event, barcode, "same") for event, barcode in evtList] + AB_list = [] + for event, barcode in evtList: + dset = list(barcodes.symmetric_difference(set([barcode]))) + idx = random.randint(0, len(dset)-1) + AB_list.append((event, dset[idx], "diff")) + + mergePairs = AA_list + AB_list + + '''读取事件、标准特征文件中数据，以 AA_list 和 AB_list 中关键字为 key 生成字典''' + results = [] + for evt, stdbcd, label in mergePairs: + ## 标准特征字典的构造方式不合适，需改进，不能用具体的barcode做key值 + + + event = evtDict[evt] + + + ## 判断是否存在轨迹图像文件夹，不存在则创建文件夹并保存轨迹图像 + pairpath = os.path.join(subimgPath, f"{evt}") + if not os.path.exists(pairpath): + os.makedirs(pairpath) + save_event_subimg(event, pairpath) + + ## 判断是否存在 barcode 样本集图像文件夹，不存在则创建文件夹并存储 barcode 样本集图像 + stdImgpath = stdDict[stdbcd]["imgpaths"] + pstdpath = os.path.join(subimgPath, f"{stdbcd}") + if not os.path.exists(pstdpath): + os.makedirs(pstdpath) + ii = 1 + for filepath in stdImgpath: + stdpath = os.path.join(pstdpath, f"{stdbcd}_{ii}.png") + shutil.copy2(filepath, stdpath) + ii += 1 + + + stdfeat = stdDict[stdbcd]["feats"] + evtfeat = event["feats_compose"] + + # 比较特征归一化前后的相似度变化情况 + + matrix = 1 - cdist(stdfeat, evtfeat, 'cosine') + simi_mean = np.mean(matrix) + simi_max = np.max(matrix) + + stdfeatm = np.mean(stdfeat, axis=0, keepdims=True) + evtfeatm = np.mean(evtfeat, axis=0, keepdims=True) + simi_mfeat = 1- np.maximum(0.0, cdist(stdfeatm, evtfeatm, 'cosine')) + + results.append((label, stdbcd, evt, simi_mean, simi_max, simi_mfeat[0,0])) + + print("contrast performance evaluate have done!") + + tm = datetime.fromtimestamp(time.time()).strftime('%Y%m%d_%H%M%S') + with open(f'{tm}.pickle', 'wb') as f: + pickle.dump(results, f) + + with open(f'{tm}.txt', 'w', encoding='utf-8') as f: + for result in results: + part = [f"{x:.3f}" if isinstance(x, float) else str(x) for x in result] + line = ', '.join(part) + f.write(line + '\n') + +def compute_contrast_accuracy(pickpath): + + pickfile = os.path.basename(pickpath) + file, _ = os.path.splitext(pickfile) + + # tm = datetime.fromtimestamp(time.time()).strftime('%Y%m%d_%H%M%S') + + + with open(pickpath, 'rb') as f: + results = pickle.load(f) + + Same, Cross = [], [] + for label, stdbcd, evt, simi_mean, simi_max, simi_mft in results: + if label == "same": + Same.append(simi_max) + if label == "diff": + Cross.append(simi_max) + + + Same = np.array(Same) + Cross = np.array(Cross) + TPFN = len(Same) + TNFP = len(Cross) + + fig, axs = plt.subplots(2, 1) + axs[0].hist(Same, bins=60, edgecolor='black') + axs[0].set_xlim([-0.2, 1]) + axs[0].set_title(f'Same Barcode, Num: {TPFN}') + + axs[1].hist(Cross, bins=60, edgecolor='black') + axs[1].set_xlim([-0.2, 1]) + axs[1].set_title(f'Cross Barcode, Num: {TNFP}') + plt.savefig(f'./result/{file}_hist.png') # svg, png, pdf + + + Recall_Pos, Recall_Neg = [], [] + Precision_Pos, Precision_Neg = [], [] + Correct = [] + Thresh = np.linspace(-0.2, 1, 100) + for th in Thresh: + TP = np.sum(Same > th) + FN = TPFN - TP + TN = np.sum(Cross < th) + FP = TNFP - TN + + Recall_Pos.append(TP/TPFN) + Recall_Neg.append(TN/TNFP) + Precision_Pos.append(TP/(TP+FP+1e-6)) + Precision_Neg.append(TN/(TN+FN+1e-6)) + Correct.append((TN+TP)/(TPFN+TNFP)) + + fig, ax = plt.subplots() + ax.plot(Thresh, Correct, 'r', label='Correct: (TN+TP)/(TPFN+TNFP)') + ax.plot(Thresh, Recall_Pos, 'b', label='Recall_Pos: TP/TPFN') + ax.plot(Thresh, Recall_Neg, 'g', label='Recall_Neg: TN/TNFP') + ax.plot(Thresh, Precision_Pos, 'c', label='Precision_Pos: TP/(TP+FP)') + ax.plot(Thresh, Precision_Neg, 'm', label='Precision_Neg: TN/(TN+FN)') + + ax.set_xlim([0, 1]) + ax.set_ylim([0, 1]) + ax.grid(True) + ax.set_title('PrecisePos & PreciseNeg') + ax.set_xlabel(f"Same Num: {TPFN}, Cross Num: {TNFP}") + ax.legend() + plt.show() + plt.savefig(f'./result/{file}_pr.png') # svg, png, pdf + + print("Haved done!!!") + + + +def generate_event_and_stdfeatures(): stdSamplePath = r"\\192.168.1.28\share\已标注数据备份\对比数据\barcode\barcode_1771" - stdBarcodePath = r"\\192.168.1.28\share\测试_202406\contrast\barcodes" - stdFeaturePath = r"\\192.168.1.28\share\测试_202406\contrast\features" + stdBarcodePath = r"\\192.168.1.28\share\测试_202406\contrast\std_barcodes" + stdFeaturePath = r"\\192.168.1.28\share\测试_202406\contrast\std_features" '''=========================== 1. 生成标准特征集 ========================''' '''1.1 提取并保存标准特征库原始图像文件地址字典''' @@ -390,46 +593,78 @@ def generate_event_and_standard_features(): # print("standard imgpath have extracted and saved") - '''1.2 特征提取，并保存至文件夹 stdFeaturePath 中''' - stdfeat_infer(stdBarcodePath, stdFeaturePath) + '''1.2 特征提取，并保存至文件夹 stdFeaturePath 中，也可在运行过程中根据barcodes交集执行''' + # stdfeat_infer(stdBarcodePath, stdFeaturePath, bcdSet=None) # print("standard features have generated!") '''=========================== 2. 提取并存储事件特征 ========================''' - shoppingEventPath = [r'\\192.168.1.28\share\测试_202406\0723\0723_1', - r'\\192.168.1.28\share\测试_202406\0723\0723_2', - r'\\192.168.1.28\share\测试_202406\0723\0723_3', - r'\\192.168.1.28\share\测试_202406\0722\0722_01', - r'\\192.168.1.28\share\测试_202406\0722\0722_02'] - shoppingFeatPath = r"\\192.168.1.28\share\测试_202406\contrast\events" - # for sPath in shoppingEventPath: - # eventList = creat_shopping_event(sPath, shoppingFeatPath) + eventDatePath = [# r'\\192.168.1.28\share\测试_202406\0723\0723_1', + # r'\\192.168.1.28\share\测试_202406\0723\0723_2', + # r'\\192.168.1.28\share\测试_202406\0723\0723_3', + # r'\\192.168.1.28\share\测试_202406\0722\0722_01', + # r'\\192.168.1.28\share\测试_202406\0722\0722_02' + # r'\\192.168.1.28\share\测试_202406\0719\719_3', + # r'\\192.168.1.28\share\测试_202406\0716\0716_1', + # r'\\192.168.1.28\share\测试_202406\0716\0716_2', + # r'\\192.168.1.28\share\测试_202406\0716\0716_3', + # r'\\192.168.1.28\share\测试_202406\0712\0712_1', # 无帧图像 + # r'\\192.168.1.28\share\测试_202406\0712\0712_2', # 无帧图像 + ] + eventFeatPath = r"\\192.168.1.28\share\测试_202406\contrast\events" + subimgPath = r'\\192.168.1.28\share\测试_202406\contrast\subimgs' + + eventList = [] + # k = 0 + for datePath in eventDatePath: + for eventName in os.listdir(datePath): + pickpath = os.path.join(eventFeatPath, f"{eventName}.pickle") + if os.path.isfile(pickpath): + continue + + eventPath = os.path.join(datePath, eventName) + + eventDict = creat_shopping_event(eventPath) + if eventDict: + eventList.append(eventDict) + with open(pickpath, 'wb') as f: + pickle.dump(eventDict, f) + print(f"Event: {eventName}, have saved!") + + # k += 1 + # if k==1: + # break + + ## 保存轨迹中 boxes 子图 + # for event in eventList: + # basename = os.path.basename(event['filepath']) + # savepath = os.path.join(subimgPath, basename) + # if not os.path.exists(savepath): + # os.makedirs(savepath) + # save_event_subimg(event, savepath) + + print("eventList have generated and features have saved!") -def shopping_event_test(): - fplist = [#r'\\192.168.1.28\share\测试_202406\0723\0723_1', - #r'\\192.168.1.28\share\测试_202406\0723\0723_2', - r'\\192.168.1.28\share\测试_202406\0723\0723_3', - #r'\\192.168.1.28\share\测试_202406\0722\0722_01', - #r'\\192.168.1.28\share\测试_202406\0722\0722_02' - ] - - shoppingFeatPath = r"\\192.168.1.28\share\测试_202406\contrast\events" - subimgPath = r'\\192.168.1.28\share\测试_202406\contrast\subimgs' - - for filepath in fplist: - '''用于保存一次购物事件的轨迹轨迹特征、及对应的图像子图''' - extract_save_trajture_subimgs(filepath, shoppingFeatPath, subimgPath) + def main(): - generate_event_and_standard_features() - # shopping_event_test() + # generate_event_and_stdfeatures() + + contrast_performance_evaluate() + + ppath = r"D:\DetectTracking\contrast" + for filename in os.listdir(ppath): + if filename.find('.pickle') < 0: + continue + + pickpath = os.path.join(ppath, filename) + compute_contrast_accuracy(pickpath) - pass if __name__ == '__main__': diff --git a/contrast/model/BAM.py b/contrast/model/BAM.py new file mode 100644 index 0000000..4ac61ae --- /dev/null +++ b/contrast/model/BAM.py @@ -0,0 +1,88 @@ +import torch.nn as nn +import torchvision +from torch.nn import init + + +class Flatten(nn.Module): + def forward(self, x): + return x.view(x.shape[0], -1) + + +class ChannelAttention(nn.Module): + def __int__(self, channel, reduction, num_layers): + super(ChannelAttention, self).__init__() + self.avgpool = nn.AdaptiveAvgPool2d(1) + gate_channels = [channel] + gate_channels += [len(channel) // reduction] * num_layers + gate_channels += [channel] + + self.ca = nn.Sequential() + self.ca.add_module('flatten', Flatten()) + for i in range(len(gate_channels) - 2): + self.ca.add_module('', nn.Linear(gate_channels[i], gate_channels[i + 1])) + self.ca.add_module('', nn.BatchNorm1d(gate_channels[i + 1])) + self.ca.add_module('', nn.ReLU()) + self.ca.add_module('', nn.Linear(gate_channels[-2], gate_channels[-1])) + + def forward(self, x): + res = self.avgpool(x) + res = self.ca(res) + res = res.unsqueeze(-1).unsqueeze(-1).expand_as(x) + return res + + +class SpatialAttention(nn.Module): + def __int__(self, channel, reduction=16, num_lay=3, dilation=2): + super(SpatialAttention).__init__() + self.sa = nn.Sequential() + self.sa.add_module('', nn.Conv2d(kernel_size=1, in_channels=channel, out_channels=(channel // reduction) * 3)) + self.sa.add_module('', nn.BatchNorm2d(num_features=(channel // reduction))) + self.sa.add_module('', nn.ReLU()) + for i in range(num_lay): + self.sa.add_module('', nn.Conv2d(kernel_size=3, + in_channels=(channel // reduction), + out_channels=(channel // reduction), + padding=1, + dilation=2)) + self.sa.add_module('', nn.BatchNorm2d(channel // reduction)) + self.sa.add_module('', nn.ReLU()) + self.sa.add_module('', nn.Conv2d(channel // reduction, 1, kernel_size=1)) + + def forward(self, x): + res = self.sa(x) + res = res.expand_as(x) + return res + + +class BAMblock(nn.Module): + def __init__(self, channel=512, reduction=16, dia_val=2): + super(BAMblock, self).__init__() + self.ca = ChannelAttention(channel, reduction) + self.sa = SpatialAttention(channel, reduction, dia_val) + self.sigmoid = nn.Sigmoid() + + def init_weights(self): + for m in self.modules(): + if isinstance(m, nn.Conv2d): + init.kaiming_normal(m.weight, mode='fan_out') + if m.bais is not None: + init.constant_(m.bias, 0) + elif isinstance(m, nn.BatchNorm2d): + init.constant_(m.weight, 1) + init.constant_(m.bias, 0) + elif isinstance(m, nn.Linear): + init.normal_(m.weight, std=0.001) + if m.bias is not None: + init.constant_(m.bias, 0) + + def forward(self, x): + b, c, _, _ = x.size() + sa_out = self.sa(x) + ca_out = self.ca(x) + weight = self.sigmoid(sa_out + ca_out) + out = (1 + weight) * x + return out + + +if __name__ == "__main__": + print(512 // 14) diff --git a/contrast/model/CBAM.py b/contrast/model/CBAM.py new file mode 100644 index 0000000..69747e0 --- /dev/null +++ b/contrast/model/CBAM.py @@ -0,0 +1,70 @@ +import torch +import torch.nn as nn +import torch.nn.init as init + +class channelAttention(nn.Module): + def __init__(self, channel, reduction=16): + super(channelAttention, self).__init__() + self.Maxpooling = nn.AdaptiveMaxPool2d(1) + self.Avepooling = nn.AdaptiveAvgPool2d(1) + self.ca = nn.Sequential() + self.ca.add_module('conv1',nn.Conv2d(channel, channel//reduction, 1, bias=False)) + self.ca.add_module('Relu', nn.ReLU()) + self.ca.add_module('conv2',nn.Conv2d(channel//reduction, channel, 1, bias=False)) + self.sigmod = nn.Sigmoid() + + def forward(self, x): + M_out = self.Maxpooling(x) + A_out = self.Avepooling(x) + M_out = self.ca(M_out) + A_out = self.ca(A_out) + out = self.sigmod(M_out+A_out) + return out + +class SpatialAttention(nn.Module): + def __init__(self, kernel_size=7): + super().__init__() + self.conv = nn.Conv2d(in_channels=2, out_channels=1, kernel_size=kernel_size, padding=kernel_size // 2) + self.sigmoid = nn.Sigmoid() + + def forward(self, x): + max_result, _ = torch.max(x, dim=1, keepdim=True) + avg_result = torch.mean(x, dim=1, keepdim=True) + result = torch.cat([max_result, avg_result], dim=1) + output = self.conv(result) + output = self.sigmoid(output) + return output + +class CBAM(nn.Module): + def __init__(self, channel, reduction=16, kernel_size=7): + super().__init__() + self.ca = channelAttention(channel, reduction) + self.sa = SpatialAttention(kernel_size) + + def init_weights(self): + for m in self.modules():#权重初始化 + if isinstance(m, nn.Conv2d): + init.kaiming_normal_(m.weight, mode='fan_out') + if m.bias is not None: + init.constant_(m.bias, 0) + elif isinstance(m, nn.BatchNorm2d): + init.constant_(m.weight, 1) + init.constant_(m.bias, 0) + elif isinstance(m, nn.Linear): + init.normal_(m.weight, std=0.001) + if m.bias is not None: + init.constant_(m.bias, 0) + + def forward(self, x): + # b,c_,_ = x.size() + # residual = x + out = x*self.ca(x) + out = out*self.sa(out) + return out + +if __name__ == '__main__': + input=torch.randn(50,512,7,7) + kernel_size=input.shape[2] + cbam = CBAM(channel=512,reduction=16,kernel_size=kernel_size) + output=cbam(input) + print(output.shape) diff --git a/contrast/model/Tool.py b/contrast/model/Tool.py new file mode 100644 index 0000000..378c11a --- /dev/null +++ b/contrast/model/Tool.py @@ -0,0 +1,33 @@ +import torch +import torch.nn as nn +import torch.nn.functional as F +class GeM(nn.Module): + def __init__(self, p=3, eps=1e-6): + super(GeM, self).__init__() + self.p = nn.Parameter(torch.ones(1) * p) + self.eps = eps + + def forward(self, x): + return self.gem(x, p=self.p, eps=self.eps, stride = 2) + + def gem(self, x, p=3, eps=1e-6, stride = 2): + return F.avg_pool2d(x.clamp(min=eps).pow(p), (x.size(-2), x.size(-1)), stride=2).pow(1. / p) + + def __repr__(self): + return self.__class__.__name__ + \ + '(' + 'p=' + '{:.4f}'.format(self.p.data.tolist()[0]) + \ + ', ' + 'eps=' + str(self.eps) + ')' + +class TripletLoss(nn.Module): + def __init__(self, margin): + super(TripletLoss, self).__init__() + self.margin = margin + + def forward(self, anchor, positive, negative, size_average = True): + distance_positive = (anchor-positive).pow(2).sum(1) + distance_negative = (anchor-negative).pow(2).sum(1) + losses = F.relu(distance_negative-distance_positive+self.margin) + return losses.mean() if size_average else losses.sum() + +if __name__ == '__main__': + print('') diff --git a/contrast/model/__init__.py b/contrast/model/__init__.py index 9eebc77..37a0635 100644 --- a/contrast/model/__init__.py +++ b/contrast/model/__init__.py @@ -1 +1,11 @@ +from .fmobilenet import FaceMobileNet +from .resnet_face import ResIRSE +from .mobilevit import mobilevit_s +from .metric import ArcFace, CosFace +from .loss import FocalLoss +from .resbam import resnet from .resnet_pre import resnet18, resnet34, resnet50, resnet14 +from .mobilenet_v2 import mobilenet_v2 +from .mobilenet_v3 import MobileNetV3_Small, MobileNetV3_Large +# from .mobilenet_v1 import mobilenet_v1 +from .lcnet import PPLCNET_x0_25, PPLCNET_x0_35, PPLCNET_x0_5, PPLCNET_x0_75, PPLCNET_x1_0, PPLCNET_x1_5, PPLCNET_x2_0, PPLCNET_x2_5 \ No newline at end of file diff --git a/contrast/model/__pycache__/BAM.cpython-38.pyc b/contrast/model/__pycache__/BAM.cpython-38.pyc new file mode 100644 index 0000000..3c10e34 Binary files /dev/null and b/contrast/model/__pycache__/BAM.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/CBAM.cpython-38.pyc b/contrast/model/__pycache__/CBAM.cpython-38.pyc new file mode 100644 index 0000000..3356083 Binary files /dev/null and b/contrast/model/__pycache__/CBAM.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/CBAM.cpython-39.pyc b/contrast/model/__pycache__/CBAM.cpython-39.pyc new file mode 100644 index 0000000..c47423e Binary files /dev/null and b/contrast/model/__pycache__/CBAM.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/Tool.cpython-38.pyc b/contrast/model/__pycache__/Tool.cpython-38.pyc new file mode 100644 index 0000000..8e85335 Binary files /dev/null and b/contrast/model/__pycache__/Tool.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/Tool.cpython-39.pyc b/contrast/model/__pycache__/Tool.cpython-39.pyc new file mode 100644 index 0000000..f009757 Binary files /dev/null and b/contrast/model/__pycache__/Tool.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/__init__.cpython-310.pyc b/contrast/model/__pycache__/__init__.cpython-310.pyc new file mode 100644 index 0000000..828ff83 Binary files /dev/null and b/contrast/model/__pycache__/__init__.cpython-310.pyc differ diff --git a/contrast/model/__pycache__/__init__.cpython-38.pyc b/contrast/model/__pycache__/__init__.cpython-38.pyc index 160bc37..653497f 100644 Binary files a/contrast/model/__pycache__/__init__.cpython-38.pyc and b/contrast/model/__pycache__/__init__.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/__init__.cpython-39.pyc b/contrast/model/__pycache__/__init__.cpython-39.pyc index 539ec67..50f8f0c 100644 Binary files a/contrast/model/__pycache__/__init__.cpython-39.pyc and b/contrast/model/__pycache__/__init__.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/fmobilenet.cpython-310.pyc b/contrast/model/__pycache__/fmobilenet.cpython-310.pyc new file mode 100644 index 0000000..2d3a370 Binary files /dev/null and b/contrast/model/__pycache__/fmobilenet.cpython-310.pyc differ diff --git a/contrast/model/__pycache__/fmobilenet.cpython-38.pyc b/contrast/model/__pycache__/fmobilenet.cpython-38.pyc new file mode 100644 index 0000000..5e2828e Binary files /dev/null and b/contrast/model/__pycache__/fmobilenet.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/fmobilenet.cpython-39.pyc b/contrast/model/__pycache__/fmobilenet.cpython-39.pyc new file mode 100644 index 0000000..c8a7647 Binary files /dev/null and b/contrast/model/__pycache__/fmobilenet.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/lcnet.cpython-38.pyc b/contrast/model/__pycache__/lcnet.cpython-38.pyc new file mode 100644 index 0000000..fa38fe4 Binary files /dev/null and b/contrast/model/__pycache__/lcnet.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/lcnet.cpython-39.pyc b/contrast/model/__pycache__/lcnet.cpython-39.pyc new file mode 100644 index 0000000..6a41ee8 Binary files /dev/null and b/contrast/model/__pycache__/lcnet.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/loss.cpython-38.pyc b/contrast/model/__pycache__/loss.cpython-38.pyc new file mode 100644 index 0000000..5aed3eb Binary files /dev/null and b/contrast/model/__pycache__/loss.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/loss.cpython-39.pyc b/contrast/model/__pycache__/loss.cpython-39.pyc new file mode 100644 index 0000000..c266cef Binary files /dev/null and b/contrast/model/__pycache__/loss.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/metric.cpython-38.pyc b/contrast/model/__pycache__/metric.cpython-38.pyc new file mode 100644 index 0000000..dc77c9d Binary files /dev/null and b/contrast/model/__pycache__/metric.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/metric.cpython-39.pyc b/contrast/model/__pycache__/metric.cpython-39.pyc new file mode 100644 index 0000000..d4a705e Binary files /dev/null and b/contrast/model/__pycache__/metric.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/mobilenet_v1.cpython-38.pyc b/contrast/model/__pycache__/mobilenet_v1.cpython-38.pyc new file mode 100644 index 0000000..8d86f36 Binary files /dev/null and b/contrast/model/__pycache__/mobilenet_v1.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/mobilenet_v1.cpython-39.pyc b/contrast/model/__pycache__/mobilenet_v1.cpython-39.pyc new file mode 100644 index 0000000..501c1e4 Binary files /dev/null and b/contrast/model/__pycache__/mobilenet_v1.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/mobilenet_v2.cpython-38.pyc b/contrast/model/__pycache__/mobilenet_v2.cpython-38.pyc new file mode 100644 index 0000000..c51a529 Binary files /dev/null and b/contrast/model/__pycache__/mobilenet_v2.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/mobilenet_v2.cpython-39.pyc b/contrast/model/__pycache__/mobilenet_v2.cpython-39.pyc new file mode 100644 index 0000000..d782cd0 Binary files /dev/null and b/contrast/model/__pycache__/mobilenet_v2.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/mobilenet_v3.cpython-38.pyc b/contrast/model/__pycache__/mobilenet_v3.cpython-38.pyc new file mode 100644 index 0000000..85471d2 Binary files /dev/null and b/contrast/model/__pycache__/mobilenet_v3.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/mobilenet_v3.cpython-39.pyc b/contrast/model/__pycache__/mobilenet_v3.cpython-39.pyc new file mode 100644 index 0000000..e1fc34a Binary files /dev/null and b/contrast/model/__pycache__/mobilenet_v3.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/mobilevit.cpython-310.pyc b/contrast/model/__pycache__/mobilevit.cpython-310.pyc new file mode 100644 index 0000000..39a75ca Binary files /dev/null and b/contrast/model/__pycache__/mobilevit.cpython-310.pyc differ diff --git a/contrast/model/__pycache__/mobilevit.cpython-38.pyc b/contrast/model/__pycache__/mobilevit.cpython-38.pyc new file mode 100644 index 0000000..b7f7051 Binary files /dev/null and b/contrast/model/__pycache__/mobilevit.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/mobilevit.cpython-39.pyc b/contrast/model/__pycache__/mobilevit.cpython-39.pyc new file mode 100644 index 0000000..aff5d93 Binary files /dev/null and b/contrast/model/__pycache__/mobilevit.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/resbam.cpython-38.pyc b/contrast/model/__pycache__/resbam.cpython-38.pyc new file mode 100644 index 0000000..e694a79 Binary files /dev/null and b/contrast/model/__pycache__/resbam.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/resbam.cpython-39.pyc b/contrast/model/__pycache__/resbam.cpython-39.pyc new file mode 100644 index 0000000..65a91de Binary files /dev/null and b/contrast/model/__pycache__/resbam.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/resnet.cpython-310.pyc b/contrast/model/__pycache__/resnet.cpython-310.pyc new file mode 100644 index 0000000..485ac48 Binary files /dev/null and b/contrast/model/__pycache__/resnet.cpython-310.pyc differ diff --git a/contrast/model/__pycache__/resnet.cpython-38.pyc b/contrast/model/__pycache__/resnet.cpython-38.pyc new file mode 100644 index 0000000..a9e8c9f Binary files /dev/null and b/contrast/model/__pycache__/resnet.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/resnet_face.cpython-38.pyc b/contrast/model/__pycache__/resnet_face.cpython-38.pyc new file mode 100644 index 0000000..118efa2 Binary files /dev/null and b/contrast/model/__pycache__/resnet_face.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/resnet_face.cpython-39.pyc b/contrast/model/__pycache__/resnet_face.cpython-39.pyc new file mode 100644 index 0000000..57209a0 Binary files /dev/null and b/contrast/model/__pycache__/resnet_face.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/resnet_pre.cpython-38.pyc b/contrast/model/__pycache__/resnet_pre.cpython-38.pyc index 199085d..786ef3a 100644 Binary files a/contrast/model/__pycache__/resnet_pre.cpython-38.pyc and b/contrast/model/__pycache__/resnet_pre.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/resnet_pre.cpython-39.pyc b/contrast/model/__pycache__/resnet_pre.cpython-39.pyc index 0301f55..d99c1c8 100644 Binary files a/contrast/model/__pycache__/resnet_pre.cpython-39.pyc and b/contrast/model/__pycache__/resnet_pre.cpython-39.pyc differ diff --git a/contrast/model/__pycache__/utils.cpython-38.pyc b/contrast/model/__pycache__/utils.cpython-38.pyc new file mode 100644 index 0000000..6e3f2bb Binary files /dev/null and b/contrast/model/__pycache__/utils.cpython-38.pyc differ diff --git a/contrast/model/__pycache__/utils.cpython-39.pyc b/contrast/model/__pycache__/utils.cpython-39.pyc new file mode 100644 index 0000000..d1836e9 Binary files /dev/null and b/contrast/model/__pycache__/utils.cpython-39.pyc differ diff --git a/contrast/model/fmobilenet.py b/contrast/model/fmobilenet.py new file mode 100644 index 0000000..2e38a44 --- /dev/null +++ b/contrast/model/fmobilenet.py @@ -0,0 +1,124 @@ +import torch +import torch.nn as nn +import torch.nn.functional as F + + + +class Flatten(nn.Module): + def forward(self, x): + return x.view(x.shape[0], -1) + +class ConvBn(nn.Module): + + def __init__(self, in_c, out_c, kernel=(1, 1), stride=1, padding=0, groups=1): + super().__init__() + self.net = nn.Sequential( + nn.Conv2d(in_c, out_c, kernel, stride, padding, groups=groups, bias=False), + nn.BatchNorm2d(out_c) + ) + + def forward(self, x): + return self.net(x) + + +class ConvBnPrelu(nn.Module): + + def __init__(self, in_c, out_c, kernel=(1, 1), stride=1, padding=0, groups=1): + super().__init__() + self.net = nn.Sequential( + ConvBn(in_c, out_c, kernel, stride, padding, groups), + nn.PReLU(out_c) + ) + + def forward(self, x): + return self.net(x) + + +class DepthWise(nn.Module): + + def __init__(self, in_c, out_c, kernel=(3, 3), stride=2, padding=1, groups=1): + super().__init__() + self.net = nn.Sequential( + ConvBnPrelu(in_c, groups, kernel=(1, 1), stride=1, padding=0), + ConvBnPrelu(groups, groups, kernel=kernel, stride=stride, padding=padding, groups=groups), + ConvBn(groups, out_c, kernel=(1, 1), stride=1, padding=0), + ) + + def forward(self, x): + return self.net(x) + + +class DepthWiseRes(nn.Module): + """DepthWise with Residual""" + + def __init__(self, in_c, out_c, kernel=(3, 3), stride=2, padding=1, groups=1): + super().__init__() + self.net = DepthWise(in_c, out_c, kernel, stride, padding, groups) + + def forward(self, x): + return self.net(x) + x + + +class MultiDepthWiseRes(nn.Module): + + def __init__(self, num_block, channels, kernel=(3, 3), stride=1, padding=1, groups=1): + super().__init__() + + self.net = nn.Sequential(*[ + DepthWiseRes(channels, channels, kernel, stride, padding, groups) + for _ in range(num_block) + ]) + + def forward(self, x): + return self.net(x) + + +class FaceMobileNet(nn.Module): + + def __init__(self, embedding_size): + super().__init__() + self.conv1 = ConvBnPrelu(1, 64, kernel=(3, 3), stride=2, padding=1) + self.conv2 = ConvBn(64, 64, kernel=(3, 3), stride=1, padding=1, groups=64) + self.conv3 = DepthWise(64, 64, kernel=(3, 3), stride=2, padding=1, groups=128) + self.conv4 = MultiDepthWiseRes(num_block=4, channels=64, kernel=3, stride=1, padding=1, groups=128) + self.conv5 = DepthWise(64, 128, kernel=(3, 3), stride=2, padding=1, groups=256) + self.conv6 = MultiDepthWiseRes(num_block=6, channels=128, kernel=(3, 3), stride=1, padding=1, groups=256) + self.conv7 = DepthWise(128, 128, kernel=(3, 3), stride=2, padding=1, groups=512) + self.conv8 = MultiDepthWiseRes(num_block=2, channels=128, kernel=(3, 3), stride=1, padding=1, groups=256) + self.conv9 = ConvBnPrelu(128, 512, kernel=(1, 1)) + self.conv10 = ConvBn(512, 512, groups=512, kernel=(7, 7)) + self.flatten = Flatten() + self.linear = nn.Linear(2048, embedding_size, bias=False) + self.bn = nn.BatchNorm1d(embedding_size) + + def forward(self, x): + #print('x',x.shape) + out = self.conv1(x) + out = self.conv2(out) + out = self.conv3(out) + out = self.conv4(out) + out = self.conv5(out) + out = self.conv6(out) + out = self.conv7(out) + out = self.conv8(out) + out = self.conv9(out) + out = self.conv10(out) + out = self.flatten(out) + out = self.linear(out) + out = self.bn(out) + return out + +if __name__ == "__main__": + from PIL import Image + import numpy as np + + x = Image.open("../samples/009.jpg").convert('L') + x = x.resize((128, 128)) + x = np.asarray(x, dtype=np.float32) + x = x[None, None, ...] + x = torch.from_numpy(x) + net = FaceMobileNet(512) + net.eval() + with torch.no_grad(): + out = net(x) + print(out.shape) diff --git a/contrast/model/lcnet.py b/contrast/model/lcnet.py new file mode 100644 index 0000000..453e8ef --- /dev/null +++ b/contrast/model/lcnet.py @@ -0,0 +1,233 @@ +import os +import torch +import torch.nn as nn +import thop + +# try: +# import softpool_cuda +# from SoftPool import soft_pool2d, SoftPool2d +# except ImportError: +# print('Please install SoftPool first: https://github.com/alexandrosstergiou/SoftPool') +# exit(0) + +NET_CONFIG = { + # k, in_c, out_c, s, use_se + "blocks2": [[3, 16, 32, 1, False]], + "blocks3": [[3, 32, 64, 2, False], [3, 64, 64, 1, False]], + "blocks4": [[3, 64, 128, 2, False], [3, 128, 128, 1, False]], + "blocks5": [[3, 128, 256, 2, False], [5, 256, 256, 1, False], + [5, 256, 256, 1, False], [5, 256, 256, 1, False], + [5, 256, 256, 1, False], [5, 256, 256, 1, False]], + "blocks6": [[5, 256, 512, 2, True], [5, 512, 512, 1, True]] +} + + +def autopad(k, p=None): + if p is None: + p = k // 2 if isinstance(k, int) else [x // 2 for x in k] + return p + + +def make_divisible(v, divisor=8, min_value=None): + if min_value is None: + min_value = divisor + new_v = max(min_value, int(v + divisor / 2) // divisor * divisor) + if new_v < 0.9 * v: + new_v += divisor + return new_v + + +class HardSwish(nn.Module): + def __init__(self, inplace=True): + super(HardSwish, self).__init__() + self.relu6 = nn.ReLU6(inplace=inplace) + + def forward(self, x): + return x * self.relu6(x+3) / 6 + + +class HardSigmoid(nn.Module): + def __init__(self, inplace=True): + super(HardSigmoid, self).__init__() + self.relu6 = nn.ReLU6(inplace=inplace) + + def forward(self, x): + return (self.relu6(x+3)) / 6 + + +class SELayer(nn.Module): + def __init__(self, channel, reduction=16): + super(SELayer, self).__init__() + self.avgpool = nn.AdaptiveAvgPool2d(1) + self.fc = nn.Sequential( + nn.Linear(channel, channel // reduction, bias=False), + nn.ReLU(inplace=True), + nn.Linear(channel // reduction, channel, bias=False), + HardSigmoid() + ) + + def forward(self, x): + b, c, h, w = x.size() + y = self.avgpool(x).view(b, c) + y = self.fc(y).view(b, c, 1, 1) + return x * y.expand_as(x) + + +class DepthwiseSeparable(nn.Module): + def __init__(self, inp, oup, dw_size, stride, use_se=False): + super(DepthwiseSeparable, self).__init__() + self.use_se = use_se + self.stride = stride + self.inp = inp + self.oup = oup + self.dw_size = dw_size + self.dw_sp = nn.Sequential( + nn.Conv2d(self.inp, self.inp, kernel_size=self.dw_size, stride=self.stride, + padding=autopad(self.dw_size, None), groups=self.inp, bias=False), + nn.BatchNorm2d(self.inp), + HardSwish(), + + nn.Conv2d(self.inp, self.oup, kernel_size=1, stride=1, padding=0, bias=False), + nn.BatchNorm2d(self.oup), + HardSwish(), + ) + self.se = SELayer(self.oup) + + def forward(self, x): + x = self.dw_sp(x) + if self.use_se: + x = self.se(x) + return x + + +class PP_LCNet(nn.Module): + def __init__(self, scale=1.0, class_num=10, class_expand=1280, dropout_prob=0.2): + super(PP_LCNet, self).__init__() + self.scale = scale + self.conv1 = nn.Conv2d(3, out_channels=make_divisible(16 * self.scale), + kernel_size=3, stride=2, padding=1, bias=False) + # k, in_c, out_c, s, use_se inp, oup, dw_size, stride, use_se=False + self.blocks2 = nn.Sequential(*[ + DepthwiseSeparable(inp=make_divisible(in_c * self.scale), + oup=make_divisible(out_c * self.scale), + dw_size=k, stride=s, use_se=use_se) + for i, (k, in_c, out_c, s, use_se) in enumerate(NET_CONFIG["blocks2"]) + ]) + + self.blocks3 = nn.Sequential(*[ + DepthwiseSeparable(inp=make_divisible(in_c * self.scale), + oup=make_divisible(out_c * self.scale), + dw_size=k, stride=s, use_se=use_se) + for i, (k, in_c, out_c, s, use_se) in enumerate(NET_CONFIG["blocks3"]) + ]) + + self.blocks4 = nn.Sequential(*[ + DepthwiseSeparable(inp=make_divisible(in_c * self.scale), + oup=make_divisible(out_c * self.scale), + dw_size=k, stride=s, use_se=use_se) + for i, (k, in_c, out_c, s, use_se) in enumerate(NET_CONFIG["blocks4"]) + ]) + # k, in_c, out_c, s, use_se inp, oup, dw_size, stride, use_se=False + self.blocks5 = nn.Sequential(*[ + DepthwiseSeparable(inp=make_divisible(in_c * self.scale), + oup=make_divisible(out_c * self.scale), + dw_size=k, stride=s, use_se=use_se) + for i, (k, in_c, out_c, s, use_se) in enumerate(NET_CONFIG["blocks5"]) + ]) + + self.blocks6 = nn.Sequential(*[ + DepthwiseSeparable(inp=make_divisible(in_c * self.scale), + oup=make_divisible(out_c * self.scale), + dw_size=k, stride=s, use_se=use_se) + for i, (k, in_c, out_c, s, use_se) in enumerate(NET_CONFIG["blocks6"]) + ]) + + self.GAP = nn.AdaptiveAvgPool2d(1) + + self.last_conv = nn.Conv2d(in_channels=make_divisible(NET_CONFIG["blocks6"][-1][2] * scale), + out_channels=class_expand, + kernel_size=1, stride=1, padding=0, bias=False) + + self.hardswish = HardSwish() + self.dropout = nn.Dropout(p=dropout_prob) + + self.fc = nn.Linear(class_expand, class_num) + + def forward(self, x): + x = self.conv1(x) + print(x.shape) + x = self.blocks2(x) + print(x.shape) + x = self.blocks3(x) + print(x.shape) + x = self.blocks4(x) + print(x.shape) + x = self.blocks5(x) + print(x.shape) + x = self.blocks6(x) + print(x.shape) + + x = self.GAP(x) + x = self.last_conv(x) + x = self.hardswish(x) + x = self.dropout(x) + x = torch.flatten(x, start_dim=1, end_dim=-1) + x = self.fc(x) + return x + + +def PPLCNET_x0_25(**kwargs): + model = PP_LCNet(scale=0.25, **kwargs) + return model + + +def PPLCNET_x0_35(**kwargs): + model = PP_LCNet(scale=0.35, **kwargs) + return model + + +def PPLCNET_x0_5(**kwargs): + model = PP_LCNet(scale=0.5, **kwargs) + return model + + +def PPLCNET_x0_75(**kwargs): + model = PP_LCNet(scale=0.75, **kwargs) + return model + + +def PPLCNET_x1_0(**kwargs): + model = PP_LCNet(scale=1.0, **kwargs) + return model + + +def PPLCNET_x1_5(**kwargs): + model = PP_LCNet(scale=1.5, **kwargs) + return model + + +def PPLCNET_x2_0(**kwargs): + model = PP_LCNet(scale=2.0, **kwargs) + return model + +def PPLCNET_x2_5(**kwargs): + model = PP_LCNet(scale=2.5, **kwargs) + return model + + + + +if __name__ == '__main__': + # input = torch.randn(1, 3, 640, 640) + # model = PPLCNET_x2_5() + # flops, params = thop.profile(model, inputs=(input,)) + # print('flops:', flops / 1000000000) + # print('params:', params / 1000000) + + model = PPLCNET_x1_0() + # model_1 = PW_Conv(3, 16) + input = torch.randn(2, 3, 256, 256) + print(input.shape) + output = model(input) + print(output.shape) # [1, num_class] + diff --git a/contrast/model/loss.py b/contrast/model/loss.py new file mode 100644 index 0000000..8f40c5c --- /dev/null +++ b/contrast/model/loss.py @@ -0,0 +1,18 @@ +import torch +import torch.nn as nn + + +class FocalLoss(nn.Module): + + def __init__(self, gamma=2): + super().__init__() + self.gamma = gamma + self.ce = torch.nn.CrossEntropyLoss() + + def forward(self, input, target): + + #print(f'theta {input.shape, input[0]}, target {target.shape, target}') + logp = self.ce(input, target) + p = torch.exp(-logp) + loss = (1 - p) ** self.gamma * logp + return loss.mean() \ No newline at end of file diff --git a/contrast/model/metric.py b/contrast/model/metric.py new file mode 100644 index 0000000..b5b0189 --- /dev/null +++ b/contrast/model/metric.py @@ -0,0 +1,83 @@ +# Definition of ArcFace loss and CosFace loss + +import math +import torch +import torch.nn as nn +import torch.nn.functional as F + + +class ArcFace(nn.Module): + + def __init__(self, embedding_size, class_num, s=30.0, m=0.50): + """ArcFace formula: + cos(m + theta) = cos(m)cos(theta) - sin(m)sin(theta) + Note that: + 0 <= m + theta <= Pi + So if (m + theta) >= Pi, then theta >= Pi - m. In [0, Pi] + we have: + cos(theta) < cos(Pi - m) + So we can use cos(Pi - m) as threshold to check whether + (m + theta) go out of [0, Pi] + + Args: + embedding_size: usually 128, 256, 512 ... + class_num: num of people when training + s: scale, see normface https://arxiv.org/abs/1704.06369 + m: margin, see SphereFace, CosFace, and ArcFace paper + """ + super().__init__() + self.in_features = embedding_size + self.out_features = class_num + self.s = s + self.m = m + self.weight = nn.Parameter(torch.FloatTensor(class_num, embedding_size)) + nn.init.xavier_uniform_(self.weight) + + self.cos_m = math.cos(m) + self.sin_m = math.sin(m) + self.th = math.cos(math.pi - m) + self.mm = math.sin(math.pi - m) * m + + def forward(self, input, label): + #print(f"embding {self.in_features}, class_num {self.out_features}, input {len(input)}, label {len(label)}") + cosine = F.linear(F.normalize(input), F.normalize(self.weight)) + # print('F.normalize(input)',input.shape) + # print('F.normalize(self.weight)',F.normalize(self.weight).shape) + sine = ((1.0 - cosine.pow(2)).clamp(0, 1)).sqrt() + phi = cosine * self.cos_m - sine * self.sin_m + phi = torch.where(cosine > self.th, phi, cosine - self.mm) # drop to CosFace + #print(f'consine {cosine.shape, cosine}, sine {sine.shape, sine}, phi {phi.shape, phi}') + # update y_i by phi in cosine + output = cosine * 1.0 # make backward works + batch_size = len(output) + output[range(batch_size), label] = phi[range(batch_size), label] + # print(f'output {(output * self.s).shape}') + # print(f'phi[range(batch_size), label] {phi[range(batch_size), label]}') + return output * self.s + + +class CosFace(nn.Module): + + def __init__(self, in_features, out_features, s=30.0, m=0.40): + """ + Args: + embedding_size: usually 128, 256, 512 ... + class_num: num of people when training + s: scale, see normface https://arxiv.org/abs/1704.06369 + m: margin, see SphereFace, CosFace, and ArcFace paper + """ + super().__init__() + self.in_features = in_features + self.out_features = out_features + self.s = s + self.m = m + self.weight = nn.Parameter(torch.FloatTensor(out_features, in_features)) + nn.init.xavier_uniform_(self.weight) + + def forward(self, input, label): + cosine = F.linear(F.normalize(input), F.normalize(self.weight)) + phi = cosine - self.m + output = cosine * 1.0 # make backward works + batch_size = len(output) + output[range(batch_size), label] = phi[range(batch_size), label] + return output * self.s \ No newline at end of file diff --git a/contrast/model/mobilenet_v1.py b/contrast/model/mobilenet_v1.py new file mode 100644 index 0000000..1262d9e --- /dev/null +++ b/contrast/model/mobilenet_v1.py @@ -0,0 +1,148 @@ +# Copyright 2022 Dakewe Biotech Corporation. All Rights Reserved. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# ============================================================================== +from typing import Callable, Any, Optional + +import torch +from torch import Tensor +from torch import nn +from torchvision.ops.misc import Conv2dNormActivation +from config import config as conf + +__all__ = [ + "MobileNetV1", + "DepthWiseSeparableConv2d", + "mobilenet_v1", +] + + +class MobileNetV1(nn.Module): + + def __init__( + self, + num_classes: int = conf.embedding_size, + ) -> None: + super(MobileNetV1, self).__init__() + self.features = nn.Sequential( + Conv2dNormActivation(3, + 32, + kernel_size=3, + stride=2, + padding=1, + norm_layer=nn.BatchNorm2d, + activation_layer=nn.ReLU, + inplace=True, + bias=False, + ), + + DepthWiseSeparableConv2d(32, 64, 1), + DepthWiseSeparableConv2d(64, 128, 2), + DepthWiseSeparableConv2d(128, 128, 1), + DepthWiseSeparableConv2d(128, 256, 2), + DepthWiseSeparableConv2d(256, 256, 1), + DepthWiseSeparableConv2d(256, 512, 2), + DepthWiseSeparableConv2d(512, 512, 1), + DepthWiseSeparableConv2d(512, 512, 1), + DepthWiseSeparableConv2d(512, 512, 1), + DepthWiseSeparableConv2d(512, 512, 1), + DepthWiseSeparableConv2d(512, 512, 1), + DepthWiseSeparableConv2d(512, 1024, 2), + DepthWiseSeparableConv2d(1024, 1024, 1), + ) + + self.avgpool = nn.AvgPool2d((7, 7)) + + self.classifier = nn.Linear(1024, num_classes) + + # Initialize neural network weights + self._initialize_weights() + + def forward(self, x: Tensor) -> Tensor: + out = self._forward_impl(x) + + return out + + # Support torch.script function + def _forward_impl(self, x: Tensor) -> Tensor: + out = self.features(x) + out = self.avgpool(out) + out = torch.flatten(out, 1) + out = self.classifier(out) + + return out + + def _initialize_weights(self) -> None: + for module in self.modules(): + if isinstance(module, nn.Conv2d): + nn.init.kaiming_normal_(module.weight, mode="fan_out", nonlinearity="relu") + if module.bias is not None: + nn.init.zeros_(module.bias) + elif isinstance(module, (nn.BatchNorm2d, nn.GroupNorm)): + nn.init.ones_(module.weight) + nn.init.zeros_(module.bias) + elif isinstance(module, nn.Linear): + nn.init.normal_(module.weight, 0, 0.01) + nn.init.zeros_(module.bias) + + +class DepthWiseSeparableConv2d(nn.Module): + def __init__( + self, + in_channels: int, + out_channels: int, + stride: int, + norm_layer: Optional[Callable[..., nn.Module]] = None + ) -> None: + super(DepthWiseSeparableConv2d, self).__init__() + self.stride = stride + if stride not in [1, 2]: + raise ValueError(f"stride should be 1 or 2 instead of {stride}") + + if norm_layer is None: + norm_layer = nn.BatchNorm2d + + self.conv = nn.Sequential( + Conv2dNormActivation(in_channels, + in_channels, + kernel_size=3, + stride=stride, + padding=1, + groups=in_channels, + norm_layer=norm_layer, + activation_layer=nn.ReLU, + inplace=True, + bias=False, + ), + Conv2dNormActivation(in_channels, + out_channels, + kernel_size=1, + stride=1, + padding=0, + norm_layer=norm_layer, + activation_layer=nn.ReLU, + inplace=True, + bias=False, + ), + + ) + + def forward(self, x: Tensor) -> Tensor: + out = self.conv(x) + + return out + + +def mobilenet_v1(**kwargs: Any) -> MobileNetV1: + model = MobileNetV1(**kwargs) + + return model diff --git a/contrast/model/mobilenet_v2.py b/contrast/model/mobilenet_v2.py new file mode 100644 index 0000000..d62f0cd --- /dev/null +++ b/contrast/model/mobilenet_v2.py @@ -0,0 +1,200 @@ +from torch import nn +from .utils import load_state_dict_from_url +from config import config as conf + +__all__ = ['MobileNetV2', 'mobilenet_v2'] + + +model_urls = { + 'mobilenet_v2': 'https://download.pytorch.org/models/mobilenet_v2-b0353104.pth', +} + + +def _make_divisible(v, divisor, min_value=None): + """ + This function is taken from the original tf repo. + It ensures that all layers have a channel number that is divisible by 8 + It can be seen here: + https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet.py + :param v: + :param divisor: + :param min_value: + :return: + """ + if min_value is None: + min_value = divisor + new_v = max(min_value, int(v + divisor / 2) // divisor * divisor) + # Make sure that round down does not go down by more than 10%. + if new_v < 0.9 * v: + new_v += divisor + return new_v + + +class ConvBNReLU(nn.Sequential): + def __init__(self, in_planes, out_planes, kernel_size=3, stride=1, groups=1, norm_layer=None): + padding = (kernel_size - 1) // 2 + if norm_layer is None: + norm_layer = nn.BatchNorm2d + super(ConvBNReLU, self).__init__( + nn.Conv2d(in_planes, out_planes, kernel_size, stride, padding, groups=groups, bias=False), + norm_layer(out_planes), + nn.ReLU6(inplace=True) + ) + + +class InvertedResidual(nn.Module): + def __init__(self, inp, oup, stride, expand_ratio, norm_layer=None): + super(InvertedResidual, self).__init__() + self.stride = stride + assert stride in [1, 2] + + if norm_layer is None: + norm_layer = nn.BatchNorm2d + + hidden_dim = int(round(inp * expand_ratio)) + self.use_res_connect = self.stride == 1 and inp == oup + + layers = [] + if expand_ratio != 1: + # pw + layers.append(ConvBNReLU(inp, hidden_dim, kernel_size=1, norm_layer=norm_layer)) + layers.extend([ + # dw + ConvBNReLU(hidden_dim, hidden_dim, stride=stride, groups=hidden_dim, norm_layer=norm_layer), + # pw-linear + nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False), + norm_layer(oup), + ]) + self.conv = nn.Sequential(*layers) + + def forward(self, x): + if self.use_res_connect: + return x + self.conv(x) + else: + return self.conv(x) + + +class MobileNetV2(nn.Module): + def __init__(self, + num_classes=conf.embedding_size, + width_mult=1.0, + inverted_residual_setting=None, + round_nearest=8, + block=None, + norm_layer=None): + """ + MobileNet V2 main class + + Args: + num_classes (int): Number of classes + width_mult (float): Width multiplier - adjusts number of channels in each layer by this amount + inverted_residual_setting: Network structure + round_nearest (int): Round the number of channels in each layer to be a multiple of this number + Set to 1 to turn off rounding + block: Module specifying inverted residual building block for mobilenet + norm_layer: Module specifying the normalization layer to use + + """ + super(MobileNetV2, self).__init__() + + if block is None: + block = InvertedResidual + + if norm_layer is None: + norm_layer = nn.BatchNorm2d + + input_channel = 32 + last_channel = 1280 + + if inverted_residual_setting is None: + inverted_residual_setting = [ + # t, c, n, s + [1, 16, 1, 1], + [6, 24, 2, 2], + [6, 32, 3, 2], + [6, 64, 4, 2], + [6, 96, 3, 1], + [6, 160, 3, 2], + [6, 320, 1, 1], + ] + + # only check the first element, assuming user knows t,c,n,s are required + if len(inverted_residual_setting) == 0 or len(inverted_residual_setting[0]) != 4: + raise ValueError("inverted_residual_setting should be non-empty " + "or a 4-element list, got {}".format(inverted_residual_setting)) + + # building first layer + input_channel = _make_divisible(input_channel * width_mult, round_nearest) + self.last_channel = _make_divisible(last_channel * max(1.0, width_mult), round_nearest) + features = [ConvBNReLU(3, input_channel, stride=2, norm_layer=norm_layer)] + # building inverted residual blocks + for t, c, n, s in inverted_residual_setting: + output_channel = _make_divisible(c * width_mult, round_nearest) + for i in range(n): + stride = s if i == 0 else 1 + features.append(block(input_channel, output_channel, stride, expand_ratio=t, norm_layer=norm_layer)) + input_channel = output_channel + # building last several layers + features.append(ConvBNReLU(input_channel, self.last_channel, kernel_size=1, norm_layer=norm_layer)) + # make it nn.Sequential + self.features = nn.Sequential(*features) + + # building classifier + self.classifier = nn.Sequential( + nn.Dropout(0.2), + nn.Linear(self.last_channel, num_classes), + ) + + # weight initialization + for m in self.modules(): + if isinstance(m, nn.Conv2d): + nn.init.kaiming_normal_(m.weight, mode='fan_out') + if m.bias is not None: + nn.init.zeros_(m.bias) + elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)): + nn.init.ones_(m.weight) + nn.init.zeros_(m.bias) + elif isinstance(m, nn.Linear): + nn.init.normal_(m.weight, 0, 0.01) + nn.init.zeros_(m.bias) + + def _forward_impl(self, x): + # This exists since TorchScript doesn't support inheritance, so the superclass method + # (this one) needs to have a name other than `forward` that can be accessed in a subclass + x = self.features(x) + # Cannot use "squeeze" as batch-size can be 1 => must use reshape with x.shape[0] + x = nn.functional.adaptive_avg_pool2d(x, 1).reshape(x.shape[0], -1) + x = self.classifier(x) + return x + + def forward(self, x): + return self._forward_impl(x) + + +def mobilenet_v2(pretrained=True, progress=True, **kwargs): + """ + Constructs a MobileNetV2 architecture from + `"MobileNetV2: Inverted Residuals and Linear Bottlenecks" `_. + + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + progress (bool): If True, displays a progress bar of the download to stderr + """ + model = MobileNetV2(**kwargs) + if pretrained: + state_dict = load_state_dict_from_url(model_urls['mobilenet_v2'], + progress=progress) + src_state_dict = state_dict + target_state_dict = model.state_dict() + skip_keys = [] + # skip mismatch size tensors in case of pretraining + for k in src_state_dict.keys(): + if k not in target_state_dict: + continue + if src_state_dict[k].size() != target_state_dict[k].size(): + skip_keys.append(k) + for k in skip_keys: + del src_state_dict[k] + missing_keys, unexpected_keys = model.load_state_dict(src_state_dict, strict=False) + #.load_state_dict(state_dict) + return model diff --git a/contrast/model/mobilenet_v3.py b/contrast/model/mobilenet_v3.py new file mode 100644 index 0000000..d69a5a0 --- /dev/null +++ b/contrast/model/mobilenet_v3.py @@ -0,0 +1,200 @@ +'''MobileNetV3 in PyTorch. + +See the paper "Inverted Residuals and Linear Bottlenecks: +Mobile Networks for Classification, Detection and Segmentation" for more details. +''' +import torch +import torch.nn as nn +import torch.nn.functional as F +from torch.nn import init +from config import config as conf + + +class hswish(nn.Module): + def forward(self, x): + out = x * F.relu6(x + 3, inplace=True) / 6 + return out + + +class hsigmoid(nn.Module): + def forward(self, x): + out = F.relu6(x + 3, inplace=True) / 6 + return out + + +class SeModule(nn.Module): + def __init__(self, in_size, reduction=4): + super(SeModule, self).__init__() + self.se = nn.Sequential( + nn.AdaptiveAvgPool2d(1), + nn.Conv2d(in_size, in_size // reduction, kernel_size=1, stride=1, padding=0, bias=False), + nn.BatchNorm2d(in_size // reduction), + nn.ReLU(inplace=True), + nn.Conv2d(in_size // reduction, in_size, kernel_size=1, stride=1, padding=0, bias=False), + nn.BatchNorm2d(in_size), + hsigmoid() + ) + + def forward(self, x): + return x * self.se(x) + + +class Block(nn.Module): + '''expand + depthwise + pointwise''' + def __init__(self, kernel_size, in_size, expand_size, out_size, nolinear, semodule, stride): + super(Block, self).__init__() + self.stride = stride + self.se = semodule + + self.conv1 = nn.Conv2d(in_size, expand_size, kernel_size=1, stride=1, padding=0, bias=False) + self.bn1 = nn.BatchNorm2d(expand_size) + self.nolinear1 = nolinear + self.conv2 = nn.Conv2d(expand_size, expand_size, kernel_size=kernel_size, stride=stride, padding=kernel_size//2, groups=expand_size, bias=False) + self.bn2 = nn.BatchNorm2d(expand_size) + self.nolinear2 = nolinear + self.conv3 = nn.Conv2d(expand_size, out_size, kernel_size=1, stride=1, padding=0, bias=False) + self.bn3 = nn.BatchNorm2d(out_size) + + self.shortcut = nn.Sequential() + if stride == 1 and in_size != out_size: + self.shortcut = nn.Sequential( + nn.Conv2d(in_size, out_size, kernel_size=1, stride=1, padding=0, bias=False), + nn.BatchNorm2d(out_size), + ) + + def forward(self, x): + out = self.nolinear1(self.bn1(self.conv1(x))) + out = self.nolinear2(self.bn2(self.conv2(out))) + out = self.bn3(self.conv3(out)) + if self.se != None: + out = self.se(out) + out = out + self.shortcut(x) if self.stride==1 else out + return out + + +class MobileNetV3_Large(nn.Module): + def __init__(self, num_classes=conf.embedding_size): + super(MobileNetV3_Large, self).__init__() + self.conv1 = nn.Conv2d(3, 16, kernel_size=3, stride=2, padding=1, bias=False) + self.bn1 = nn.BatchNorm2d(16) + self.hs1 = hswish() + + self.bneck = nn.Sequential( + Block(3, 16, 16, 16, nn.ReLU(inplace=True), None, 1), + Block(3, 16, 64, 24, nn.ReLU(inplace=True), None, 2), + Block(3, 24, 72, 24, nn.ReLU(inplace=True), None, 1), + Block(5, 24, 72, 40, nn.ReLU(inplace=True), SeModule(40), 2), + Block(5, 40, 120, 40, nn.ReLU(inplace=True), SeModule(40), 1), + Block(5, 40, 120, 40, nn.ReLU(inplace=True), SeModule(40), 1), + Block(3, 40, 240, 80, hswish(), None, 2), + Block(3, 80, 200, 80, hswish(), None, 1), + Block(3, 80, 184, 80, hswish(), None, 1), + Block(3, 80, 184, 80, hswish(), None, 1), + Block(3, 80, 480, 112, hswish(), SeModule(112), 1), + Block(3, 112, 672, 112, hswish(), SeModule(112), 1), + Block(5, 112, 672, 160, hswish(), SeModule(160), 1), + Block(5, 160, 672, 160, hswish(), SeModule(160), 2), + Block(5, 160, 960, 160, hswish(), SeModule(160), 1), + ) + + + self.conv2 = nn.Conv2d(160, 960, kernel_size=1, stride=1, padding=0, bias=False) + self.bn2 = nn.BatchNorm2d(960) + self.hs2 = hswish() + self.linear3 = nn.Linear(960, 1280) + self.bn3 = nn.BatchNorm1d(1280) + self.hs3 = hswish() + self.linear4 = nn.Linear(1280, num_classes) + self.init_params() + + def init_params(self): + for m in self.modules(): + if isinstance(m, nn.Conv2d): + init.kaiming_normal_(m.weight, mode='fan_out') + if m.bias is not None: + init.constant_(m.bias, 0) + elif isinstance(m, nn.BatchNorm2d): + init.constant_(m.weight, 1) + init.constant_(m.bias, 0) + elif isinstance(m, nn.Linear): + init.normal_(m.weight, std=0.001) + if m.bias is not None: + init.constant_(m.bias, 0) + + def forward(self, x): + out = self.hs1(self.bn1(self.conv1(x))) + out = self.bneck(out) + out = self.hs2(self.bn2(self.conv2(out))) + out = F.avg_pool2d(out, conf.img_size // 32) + out = out.view(out.size(0), -1) + out = self.hs3(self.bn3(self.linear3(out))) + out = self.linear4(out) + return out + + + +class MobileNetV3_Small(nn.Module): + def __init__(self, num_classes=conf.embedding_size): + super(MobileNetV3_Small, self).__init__() + self.conv1 = nn.Conv2d(3, 16, kernel_size=3, stride=2, padding=1, bias=False) + self.bn1 = nn.BatchNorm2d(16) + self.hs1 = hswish() + + self.bneck = nn.Sequential( + Block(3, 16, 16, 16, nn.ReLU(inplace=True), SeModule(16), 2), + Block(3, 16, 72, 24, nn.ReLU(inplace=True), None, 2), + Block(3, 24, 88, 24, nn.ReLU(inplace=True), None, 1), + Block(5, 24, 96, 40, hswish(), SeModule(40), 2), + Block(5, 40, 240, 40, hswish(), SeModule(40), 1), + Block(5, 40, 240, 40, hswish(), SeModule(40), 1), + Block(5, 40, 120, 48, hswish(), SeModule(48), 1), + Block(5, 48, 144, 48, hswish(), SeModule(48), 1), + Block(5, 48, 288, 96, hswish(), SeModule(96), 2), + Block(5, 96, 576, 96, hswish(), SeModule(96), 1), + Block(5, 96, 576, 96, hswish(), SeModule(96), 1), + ) + + + self.conv2 = nn.Conv2d(96, 576, kernel_size=1, stride=1, padding=0, bias=False) + self.bn2 = nn.BatchNorm2d(576) + self.hs2 = hswish() + self.linear3 = nn.Linear(576, 1280) + self.bn3 = nn.BatchNorm1d(1280) + self.hs3 = hswish() + self.linear4 = nn.Linear(1280, num_classes) + self.init_params() + + def init_params(self): + for m in self.modules(): + if isinstance(m, nn.Conv2d): + init.kaiming_normal_(m.weight, mode='fan_out') + if m.bias is not None: + init.constant_(m.bias, 0) + elif isinstance(m, nn.BatchNorm2d): + init.constant_(m.weight, 1) + init.constant_(m.bias, 0) + elif isinstance(m, nn.Linear): + init.normal_(m.weight, std=0.001) + if m.bias is not None: + init.constant_(m.bias, 0) + + def forward(self, x): + out = self.hs1(self.bn1(self.conv1(x))) + out = self.bneck(out) + out = self.hs2(self.bn2(self.conv2(out))) + out = F.avg_pool2d(out, conf.img_size // 32) + out = out.view(out.size(0), -1) + + out = self.hs3(self.bn3(self.linear3(out))) + out = self.linear4(out) + return out + + + +def test(): + net = MobileNetV3_Small() + x = torch.randn(2,3,224,224) + y = net(x) + print(y.size()) + +# test() \ No newline at end of file diff --git a/contrast/model/mobilevit.py b/contrast/model/mobilevit.py new file mode 100644 index 0000000..f371ee9 --- /dev/null +++ b/contrast/model/mobilevit.py @@ -0,0 +1,265 @@ +import torch +import torch.nn as nn + +from einops import rearrange +from config import config as conf + + +def conv_1x1_bn(inp, oup): + return nn.Sequential( + nn.Conv2d(inp, oup, 1, 1, 0, bias=False), + nn.BatchNorm2d(oup), + nn.SiLU() + ) + + +def conv_nxn_bn(inp, oup, kernal_size=3, stride=1): + return nn.Sequential( + nn.Conv2d(inp, oup, kernal_size, stride, 1, bias=False), + nn.BatchNorm2d(oup), + nn.SiLU() + ) + + +class PreNorm(nn.Module): + def __init__(self, dim, fn): + super().__init__() + self.norm = nn.LayerNorm(dim) + self.fn = fn + + def forward(self, x, **kwargs): + return self.fn(self.norm(x), **kwargs) + + +class FeedForward(nn.Module): + def __init__(self, dim, hidden_dim, dropout=0.): + super().__init__() + self.net = nn.Sequential( + nn.Linear(dim, hidden_dim), + nn.SiLU(), + nn.Dropout(dropout), + nn.Linear(hidden_dim, dim), + nn.Dropout(dropout) + ) + + def forward(self, x): + return self.net(x) + + +class Attention(nn.Module): + def __init__(self, dim, heads=8, dim_head=64, dropout=0.): + super().__init__() + inner_dim = dim_head * heads + project_out = not (heads == 1 and dim_head == dim) + + self.heads = heads + self.scale = dim_head ** -0.5 + + self.attend = nn.Softmax(dim=-1) + self.to_qkv = nn.Linear(dim, inner_dim * 3, bias=False) + + self.to_out = nn.Sequential( + nn.Linear(inner_dim, dim), + nn.Dropout(dropout) + ) if project_out else nn.Identity() + + def forward(self, x): + qkv = self.to_qkv(x).chunk(3, dim=-1) + q, k, v = map(lambda t: rearrange(t, 'b p n (h d) -> b p h n d', h=self.heads), qkv) + + dots = torch.matmul(q, k.transpose(-1, -2)) * self.scale + attn = self.attend(dots) + out = torch.matmul(attn, v) + out = rearrange(out, 'b p h n d -> b p n (h d)') + return self.to_out(out) + + +class Transformer(nn.Module): + def __init__(self, dim, depth, heads, dim_head, mlp_dim, dropout=0.): + super().__init__() + self.layers = nn.ModuleList([]) + for _ in range(depth): + self.layers.append(nn.ModuleList([ + PreNorm(dim, Attention(dim, heads, dim_head, dropout)), + PreNorm(dim, FeedForward(dim, mlp_dim, dropout)) + ])) + + def forward(self, x): + for attn, ff in self.layers: + x = attn(x) + x + x = ff(x) + x + return x + + +class MV2Block(nn.Module): + def __init__(self, inp, oup, stride=1, expansion=4): + super().__init__() + self.stride = stride + assert stride in [1, 2] + + hidden_dim = int(inp * expansion) + self.use_res_connect = self.stride == 1 and inp == oup + + if expansion == 1: + self.conv = nn.Sequential( + # dw + nn.Conv2d(hidden_dim, hidden_dim, 3, stride, 1, groups=hidden_dim, bias=False), + nn.BatchNorm2d(hidden_dim), + nn.SiLU(), + # pw-linear + nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False), + nn.BatchNorm2d(oup), + ) + else: + self.conv = nn.Sequential( + # pw + nn.Conv2d(inp, hidden_dim, 1, 1, 0, bias=False), + nn.BatchNorm2d(hidden_dim), + nn.SiLU(), + # dw + nn.Conv2d(hidden_dim, hidden_dim, 3, stride, 1, groups=hidden_dim, bias=False), + nn.BatchNorm2d(hidden_dim), + nn.SiLU(), + # pw-linear + nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False), + nn.BatchNorm2d(oup), + ) + + def forward(self, x): + if self.use_res_connect: + return x + self.conv(x) + else: + return self.conv(x) + + +class MobileViTBlock(nn.Module): + def __init__(self, dim, depth, channel, kernel_size, patch_size, mlp_dim, dropout=0.): + super().__init__() + self.ph, self.pw = patch_size + + self.conv1 = conv_nxn_bn(channel, channel, kernel_size) + self.conv2 = conv_1x1_bn(channel, dim) + + self.transformer = Transformer(dim, depth, 4, 8, mlp_dim, dropout) + + self.conv3 = conv_1x1_bn(dim, channel) + self.conv4 = conv_nxn_bn(2 * channel, channel, kernel_size) + + def forward(self, x): + y = x.clone() + + # Local representations + x = self.conv1(x) + x = self.conv2(x) + + # Global representations + _, _, h, w = x.shape + x = rearrange(x, 'b d (h ph) (w pw) -> b (ph pw) (h w) d', ph=self.ph, pw=self.pw) + x = self.transformer(x) + x = rearrange(x, 'b (ph pw) (h w) d -> b d (h ph) (w pw)', h=h // self.ph, w=w // self.pw, ph=self.ph, + pw=self.pw) + + # Fusion + x = self.conv3(x) + x = torch.cat((x, y), 1) + x = self.conv4(x) + return x + + +class MobileViT(nn.Module): + def __init__(self, image_size, dims, channels, num_classes, expansion=4, kernel_size=3, patch_size=(2, 2)): + super().__init__() + ih, iw = image_size + ph, pw = patch_size + assert ih % ph == 0 and iw % pw == 0 + + L = [2, 4, 3] + + self.conv1 = conv_nxn_bn(3, channels[0], stride=2) + + self.mv2 = nn.ModuleList([]) + self.mv2.append(MV2Block(channels[0], channels[1], 1, expansion)) + self.mv2.append(MV2Block(channels[1], channels[2], 2, expansion)) + self.mv2.append(MV2Block(channels[2], channels[3], 1, expansion)) + self.mv2.append(MV2Block(channels[2], channels[3], 1, expansion)) # Repeat + self.mv2.append(MV2Block(channels[3], channels[4], 2, expansion)) + self.mv2.append(MV2Block(channels[5], channels[6], 2, expansion)) + self.mv2.append(MV2Block(channels[7], channels[8], 2, expansion)) + + self.mvit = nn.ModuleList([]) + self.mvit.append(MobileViTBlock(dims[0], L[0], channels[5], kernel_size, patch_size, int(dims[0] * 2))) + self.mvit.append(MobileViTBlock(dims[1], L[1], channels[7], kernel_size, patch_size, int(dims[1] * 4))) + self.mvit.append(MobileViTBlock(dims[2], L[2], channels[9], kernel_size, patch_size, int(dims[2] * 4))) + + self.conv2 = conv_1x1_bn(channels[-2], channels[-1]) + + self.pool = nn.AvgPool2d(ih // 32, 1) + self.fc = nn.Linear(channels[-1], num_classes, bias=False) + + def forward(self, x): + #print('x',x.shape) + x = self.conv1(x) + x = self.mv2[0](x) + + x = self.mv2[1](x) + x = self.mv2[2](x) + x = self.mv2[3](x) # Repeat + + x = self.mv2[4](x) + x = self.mvit[0](x) + + x = self.mv2[5](x) + x = self.mvit[1](x) + + x = self.mv2[6](x) + x = self.mvit[2](x) + x = self.conv2(x) + + + #print('pool_before',x.shape) + x = self.pool(x).view(-1, x.shape[1]) + #print('self_pool',self.pool) + #print('pool_after',x.shape) + x = self.fc(x) + return x + + +def mobilevit_xxs(): + dims = [64, 80, 96] + channels = [16, 16, 24, 24, 48, 48, 64, 64, 80, 80, 320] + return MobileViT((256, 256), dims, channels, num_classes=1000, expansion=2) + + +def mobilevit_xs(): + dims = [96, 120, 144] + channels = [16, 32, 48, 48, 64, 64, 80, 80, 96, 96, 384] + return MobileViT((256, 256), dims, channels, num_classes=1000) + + +def mobilevit_s(): + dims = [144, 192, 240] + channels = [16, 32, 64, 64, 96, 96, 128, 128, 160, 160, 640] + return MobileViT((conf.img_size, conf.img_size), dims, channels, num_classes=conf.embedding_size) + + +def count_parameters(model): + return sum(p.numel() for p in model.parameters() if p.requires_grad) + + +if __name__ == '__main__': + img = torch.randn(5, 3, 256, 256) + + vit = mobilevit_xxs() + out = vit(img) + print(out.shape) + print(count_parameters(vit)) + + vit = mobilevit_xs() + out = vit(img) + print(out.shape) + print(count_parameters(vit)) + + vit = mobilevit_s() + out = vit(img) + print(out.shape) + print(count_parameters(vit)) diff --git a/contrast/model/resbam.py b/contrast/model/resbam.py new file mode 100644 index 0000000..21395c3 --- /dev/null +++ b/contrast/model/resbam.py @@ -0,0 +1,142 @@ +from model.CBAM import CBAM +import torch +import torch.nn as nn +from model.Tool import GeM as gem + + +class Bottleneck(nn.Module): + expansion = 4 + + def __init__(self, inchannel, outchannel, stride=1, dowsample=None): + # super(Bottleneck, self).__init__() + super().__init__() + self.conv1 = nn.Conv2d(in_channels=inchannel, out_channels=outchannel, kernel_size=1, stride=1, bias=False) + self.bn1 = nn.BatchNorm2d(outchannel) + self.conv2 = nn.Conv2d(in_channels=outchannel, out_channels=outchannel, kernel_size=3, bias=False, + stride=stride, padding=1) + self.bn2 = nn.BatchNorm2d(outchannel) + self.conv3 = nn.Conv2d(in_channels=outchannel, out_channels=outchannel * self.expansion, stride=1, bias=False, + kernel_size=1) + self.bn3 = nn.BatchNorm2d(outchannel * self.expansion) + self.relu = nn.ReLU(inplace=True) + self.downsample = dowsample + + def forward(self, x): + self.identity = x + # print('>>>>>>>>',type(x)) + if self.downsample is not None: + # print('>>>>downsample>>>>', type(self.downsample)) + self.identity = self.downsample(x) + out = self.conv1(x) + out = self.bn1(out) + out = self.relu(out) + out = self.conv2(out) + out = self.bn2(out) + out = self.relu(out) + out = self.conv3(out) + out = self.bn3(out) + # print('>>>>out>>>identity',out.size(),self.identity.size()) + out = out + self.identity + out = self.relu(out) + return out + + +class resnet(nn.Module): + def __init__(self, block=Bottleneck, block_num=[3, 4, 6, 3], num_class=1000): + super().__init__() + self.in_channel = 64 + self.conv1 = nn.Conv2d(in_channels=3, + out_channels=self.in_channel, + stride=2, + kernel_size=7, + padding=3, + bias=False) + self.bn1 = nn.BatchNorm2d(self.in_channel) + self.relu = nn.ReLU(inplace=True) + self.cbam = CBAM(self.in_channel) + self.cbam1 = CBAM(2048) + self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) + self.layer1 = self._make_layer(block, 64, block_num[0], stride=1) + self.layer2 = self._make_layer(block, 128, block_num[1], stride=2) + self.layer3 = self._make_layer(block, 256, block_num[2], stride=2) + self.layer4 = self._make_layer(block, 512, block_num[3], stride=2) + self.avgpool = nn.AdaptiveAvgPool2d((1, 1)) + self.gem = gem() + self.fc = nn.Linear(512 * block.expansion, num_class) + + for m in self.modules(): + if isinstance(m, nn.Conv2d): + nn.init.kaiming_normal(m.weight, mode='fan_out', + nonlinearity='relu') + if isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)): + nn.init.constant_(m.weight, 1.0) + nn.init.constant_(m.bias, 1.0) + + def _make_layer(self, block, channel, block_num, stride=1): + downsample = None + if stride != 1 or self.in_channel != channel * block.expansion: + downsample = nn.Sequential( + nn.Conv2d(self.in_channel, channel * block.expansion, kernel_size=1, stride=stride, bias=False), + nn.BatchNorm2d(channel * block.expansion)) + layer = [] + layer.append(block(self.in_channel, channel, stride, downsample)) + self.in_channel = channel * block.expansion + for _ in range(1, block_num): + layer.append(block(self.in_channel, channel)) + return nn.Sequential(*layer) + + def forward(self, x): + x = self.conv1(x) + x = self.bn1(x) + x = self.relu(x) + x = self.maxpool(x) + x = self.cbam(x) + + x = self.layer1(x) + x = self.layer2(x) + x = self.layer3(x) + x = self.layer4(x) + + x = self.cbam1(x) + # x = self.avgpool(x) + x = self.gem(x) + x = torch.flatten(x, 1) + x = self.fc(x) + return x + + +class TripletNet(nn.Module): + def __init__(self, num_class, flag=True): + super(TripletNet, self).__init__() + self.initnet = rescbam(num_class) + self.flag = flag + + def forward(self, x1, x2=None, x3=None): + if self.flag: + output1 = self.initnet(x1) + output2 = self.initnet(x2) + output3 = self.initnet(x3) + return output1, output2, output3 + else: + output = self.initnet(x1) + return output + + +def rescbam(num_class): + return resnet(block=Bottleneck, block_num=[3, 4, 6, 3], num_class=num_class) + + +if __name__ == '__main__': + input1 = torch.randn(4, 3, 640, 640) + input2 = torch.randn(4, 3, 640, 640) + input3 = torch.randn(4, 3, 640, 640) + + # rescbam测试 + # Resnet50 = rescbam(512) + # output = Resnet50.forward(input1) + # print(Resnet50) + + # trnet测试 + trnet = TripletNet(512) + output = trnet(input1, input2, input3) + print(output) diff --git a/contrast/model/resnet.py b/contrast/model/resnet.py new file mode 100644 index 0000000..3c4fdf0 --- /dev/null +++ b/contrast/model/resnet.py @@ -0,0 +1,189 @@ +"""resnet in pytorch + + + +[1] Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun. + + Deep Residual Learning for Image Recognition + https://arxiv.org/abs/1512.03385v1 +""" + +import torch +import torch.nn as nn +from config import config as conf +from CBAM import CBAM + +class BasicBlock(nn.Module): + """Basic Block for resnet 18 and resnet 34 + + """ + + #BasicBlock and BottleNeck block + #have different output size + #we use class attribute expansion + #to distinct + expansion = 1 + + def __init__(self, in_channels, out_channels, stride=1): + super().__init__() + + #residual function + self.residual_function = nn.Sequential( + nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=stride, padding=1, bias=False), + nn.BatchNorm2d(out_channels), + nn.ReLU(inplace=True), + nn.Conv2d(out_channels, out_channels * BasicBlock.expansion, kernel_size=3, padding=1, bias=False), + nn.BatchNorm2d(out_channels * BasicBlock.expansion) + ) + + #shortcut + self.shortcut = nn.Sequential() + + #the shortcut output dimension is not the same with residual function + #use 1*1 convolution to match the dimension + if stride != 1 or in_channels != BasicBlock.expansion * out_channels: + self.shortcut = nn.Sequential( + nn.Conv2d(in_channels, out_channels * BasicBlock.expansion, kernel_size=1, stride=stride, bias=False), + nn.BatchNorm2d(out_channels * BasicBlock.expansion) + ) + + def forward(self, x): + return nn.ReLU(inplace=True)(self.residual_function(x) + self.shortcut(x)) + +class BottleNeck(nn.Module): + """Residual block for resnet over 50 layers + + """ + expansion = 4 + def __init__(self, in_channels, out_channels, stride=1): + super().__init__() + self.residual_function = nn.Sequential( + nn.Conv2d(in_channels, out_channels, kernel_size=1, bias=False), + nn.BatchNorm2d(out_channels), + nn.ReLU(inplace=True), + nn.Conv2d(out_channels, out_channels, stride=stride, kernel_size=3, padding=1, bias=False), + nn.BatchNorm2d(out_channels), + nn.ReLU(inplace=True), + nn.Conv2d(out_channels, out_channels * BottleNeck.expansion, kernel_size=1, bias=False), + nn.BatchNorm2d(out_channels * BottleNeck.expansion), + ) + + self.shortcut = nn.Sequential() + + if stride != 1 or in_channels != out_channels * BottleNeck.expansion: + self.shortcut = nn.Sequential( + nn.Conv2d(in_channels, out_channels * BottleNeck.expansion, stride=stride, kernel_size=1, bias=False), + nn.BatchNorm2d(out_channels * BottleNeck.expansion) + ) + + def forward(self, x): + return nn.ReLU(inplace=True)(self.residual_function(x) + self.shortcut(x)) + +class ResNet(nn.Module): + + def __init__(self, block, num_block, cbam = False, num_classes=conf.embedding_size): + super().__init__() + + self.in_channels = 64 + + # self.conv1 = nn.Sequential( + # nn.Conv2d(3, 64, kernel_size=3, padding=1, bias=False), + # nn.BatchNorm2d(64), + # nn.ReLU(inplace=True)) + + self.conv1 = nn.Sequential( + nn.Conv2d(3, 64,stride=2,kernel_size=7,padding=3,bias=False), + nn.BatchNorm2d(64), + nn.ReLU(inplace=True), + nn.MaxPool2d(kernel_size=3, stride=2, padding=1)) + + self.cbam = CBAM(self.in_channels) + + #we use a different inputsize than the original paper + #so conv2_x's stride is 1 + self.conv2_x = self._make_layer(block, 64, num_block[0], 1) + self.conv3_x = self._make_layer(block, 128, num_block[1], 2) + self.conv4_x = self._make_layer(block, 256, num_block[2], 2) + self.conv5_x = self._make_layer(block, 512, num_block[3], 2) + self.cbam1 = CBAM(self.in_channels) + self.avg_pool = nn.AdaptiveAvgPool2d((1, 1)) + self.fc = nn.Linear(512 * block.expansion, num_classes) + + for m in self.modules(): + if isinstance(m, nn.Conv2d): + nn.init.kaiming_normal(m.weight,mode = 'fan_out', + nonlinearity='relu') + if isinstance(m, (nn.BatchNorm2d)): + nn.init.constant_(m.weight, 1.0) + nn.init.constant_(m.bias, 1.0) + + def _make_layer(self, block, out_channels, num_blocks, stride): + """make resnet layers(by layer i didnt mean this 'layer' was the + same as a neuron netowork layer, ex. conv layer), one layer may + contain more than one residual block + + Args: + block: block type, basic block or bottle neck block + out_channels: output depth channel number of this layer + num_blocks: how many blocks per layer + stride: the stride of the first block of this layer + + Return: + return a resnet layer + """ + + # we have num_block blocks per layer, the first block + # could be 1 or 2, other blocks would always be 1 + strides = [stride] + [1] * (num_blocks - 1) + layers = [] + for stride in strides: + layers.append(block(self.in_channels, out_channels, stride)) + self.in_channels = out_channels * block.expansion + + return nn.Sequential(*layers) + + def forward(self, x): + output = self.conv1(x) + if cbam: + output = self.cbam(x) + output = self.conv2_x(output) + output = self.conv3_x(output) + output = self.conv4_x(output) + output = self.conv5_x(output) + if cbam: + output = self.cbam1(x) + print('pollBefore',output.shape) + output = self.avg_pool(output) + print('poolAfter',output.shape) + output = output.view(output.size(0), -1) + print('fcBefore',output.shape) + output = self.fc(output) + + return output + +def resnet18(cbam = False): + """ return a ResNet 18 object + """ + return ResNet(BasicBlock, [2, 2, 2, 2], cbam) + +def resnet34(): + """ return a ResNet 34 object + """ + return ResNet(BasicBlock, [3, 4, 6, 3]) + +def resnet50(): + """ return a ResNet 50 object + """ + return ResNet(BottleNeck, [3, 4, 6, 3]) + +def resnet101(): + """ return a ResNet 101 object + """ + return ResNet(BottleNeck, [3, 4, 23, 3]) + +def resnet152(): + """ return a ResNet 152 object + """ + return ResNet(BottleNeck, [3, 8, 36, 3]) + + diff --git a/contrast/model/resnet_face.py b/contrast/model/resnet_face.py new file mode 100644 index 0000000..86f1b2b --- /dev/null +++ b/contrast/model/resnet_face.py @@ -0,0 +1,121 @@ +""" Resnet_IR_SE in ArcFace """ + +import torch +import torch.nn as nn +import torch.nn.functional as F + + +class Flatten(nn.Module): + def forward(self, x): + return x.reshape(x.shape[0], -1) + + +class SEConv(nn.Module): + """Use Convolution instead of FullyConnection in SE""" + + def __init__(self, channels, reduction): + super().__init__() + self.net = nn.Sequential( + nn.AdaptiveAvgPool2d(1), + nn.Conv2d(channels, channels // reduction, kernel_size=1, bias=False), + nn.ReLU(inplace=True), + nn.Conv2d(channels // reduction, channels, kernel_size=1, bias=False), + nn.Sigmoid(), + ) + + def forward(self, x): + return self.net(x) * x + + +class SE(nn.Module): + + def __init__(self, channels, reduction): + super().__init__() + self.net = nn.Sequential( + nn.AdaptiveAvgPool2d(1), + nn.Linear(channels, channels // reduction), + nn.ReLU(inplace=True), + nn.Linear(channels // reduction, channels), + nn.Sigmoid(), + ) + + def forward(self, x): + return self.net(x) * x + + +class IRSE(nn.Module): + + def __init__(self, channels, depth, stride): + super().__init__() + if channels == depth: + self.shortcut = nn.MaxPool2d(kernel_size=1, stride=stride) + else: + self.shortcut = nn.Sequential( + nn.Conv2d(channels, depth, (1, 1), stride, bias=False), + nn.BatchNorm2d(depth), + ) + self.residual = nn.Sequential( + nn.BatchNorm2d(channels), + nn.Conv2d(channels, depth, (3, 3), 1, 1, bias=False), + nn.PReLU(depth), + nn.Conv2d(depth, depth, (3, 3), stride, 1, bias=False), + nn.BatchNorm2d(depth), + SEConv(depth, 16), + ) + + def forward(self, x): + return self.shortcut(x) + self.residual(x) + + +class ResIRSE(nn.Module): + """Resnet50-IRSE backbone""" + + def __init__(self, ih, embedding_size, drop_ratio): + super().__init__() + ih_last = ih // 16 + self.input_layer = nn.Sequential( + nn.Conv2d(3, 64, (3, 3), 1, 1, bias=False), + nn.BatchNorm2d(64), + nn.PReLU(64), + ) + self.output_layer = nn.Sequential( + nn.BatchNorm2d(512), + nn.Dropout(drop_ratio), + Flatten(), + nn.Linear(512 * ih_last * ih_last, embedding_size), + nn.BatchNorm1d(embedding_size), + ) + + # ["channels", "depth", "stride"], + self.res50_arch = [ + [64, 64, 2], [64, 64, 1], [64, 64, 1], + [64, 128, 2], [128, 128, 1], [128, 128, 1], [128, 128, 1], + [128, 256, 2], [256, 256, 1], [256, 256, 1], [256, 256, 1], [256, 256, 1], + [256, 256, 1], [256, 256, 1], [256, 256, 1], [256, 256, 1], [256, 256, 1], + [256, 256, 1], [256, 256, 1], [256, 256, 1], [256, 256, 1], + [256, 512, 2], [512, 512, 1], [512, 512, 1], + ] + + self.body = nn.Sequential(*[IRSE(a, b, c) for (a, b, c) in self.res50_arch]) + + def forward(self, x): + x = self.input_layer(x) + x = self.body(x) + x = self.output_layer(x) + return x + + +if __name__ == "__main__": + from PIL import Image + import numpy as np + + x = Image.open("../samples/009.jpg").convert('L') + x = x.resize((128, 128)) + x = np.asarray(x, dtype=np.float32) + x = x[None, None, ...] + x = torch.from_numpy(x) + net = ResIRSE(512, 0.6) + net.eval() + with torch.no_grad(): + out = net(x) + print(out.shape) diff --git a/contrast/model/utils.py b/contrast/model/utils.py new file mode 100644 index 0000000..638ef07 --- /dev/null +++ b/contrast/model/utils.py @@ -0,0 +1,4 @@ +try: + from torch.hub import load_state_dict_from_url +except ImportError: + from torch.utils.model_zoo import load_url as load_state_dict_from_url diff --git a/contrast/model/vit.py b/contrast/model/vit.py new file mode 100644 index 0000000..a6231f1 --- /dev/null +++ b/contrast/model/vit.py @@ -0,0 +1,137 @@ +import torch +from vit_pytorch.mobile_vit import MobileViT +from vit_pytorch import vit +from vit_pytorch import SimpleViT +import torch +from torch import nn + +from einops import rearrange, repeat +from einops.layers.torch import Rearrange + + +# helpers + +def pair(t): + return t if isinstance(t, tuple) else (t, t) + + +# classes + +class FeedForward(nn.Module): + def __init__(self, dim, hidden_dim, dropout=0.): + super().__init__() + self.net = nn.Sequential( + nn.LayerNorm(dim), + nn.Linear(dim, hidden_dim), + nn.GELU(), + nn.Dropout(dropout), + nn.Linear(hidden_dim, dim), + nn.Dropout(dropout) + ) + + def forward(self, x): + return self.net(x) + + +class Attention(nn.Module): + def __init__(self, dim, heads=8, dim_head=64, dropout=0.): + super().__init__() + inner_dim = dim_head * heads + project_out = not (heads == 1 and dim_head == dim) + + self.heads = heads + self.scale = dim_head ** -0.5 + + self.norm = nn.LayerNorm(dim) + + self.attend = nn.Softmax(dim=-1) + self.dropout = nn.Dropout(dropout) + + self.to_qkv = nn.Linear(dim, inner_dim * 3, bias=False) + + self.to_out = nn.Sequential( + nn.Linear(inner_dim, dim), + nn.Dropout(dropout) + ) if project_out else nn.Identity() + + def forward(self, x): + x = self.norm(x) + + qkv = self.to_qkv(x).chunk(3, dim=-1) + q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> b h n d', h=self.heads), qkv) + + dots = torch.matmul(q, k.transpose(-1, -2)) * self.scale + + attn = self.attend(dots) + attn = self.dropout(attn) + + out = torch.matmul(attn, v) + out = rearrange(out, 'b h n d -> b n (h d)') + return self.to_out(out) + + +class Transformer(nn.Module): + def __init__(self, dim, depth, heads, dim_head, mlp_dim, dropout=0.): + super().__init__() + self.norm = nn.LayerNorm(dim) + self.layers = nn.ModuleList([]) + for _ in range(depth): + self.layers.append(nn.ModuleList([ + Attention(dim, heads=heads, dim_head=dim_head, dropout=dropout), + FeedForward(dim, mlp_dim, dropout=dropout) + ])) + + def forward(self, x): + for attn, ff in self.layers: + x = attn(x) + x + x = ff(x) + x + + return self.norm(x) + + +class ViT(nn.Module): + def __init__(self, *, image_size, patch_size, num_classes, dim, depth, heads, mlp_dim, pool='cls', channels=3, + dim_head=64, dropout=0., emb_dropout=0.): + super().__init__() + image_height, image_width = pair(image_size) + patch_height, patch_width = pair(patch_size) + + assert image_height % patch_height == 0 and image_width % patch_width == 0, 'Image dimensions must be divisible by the patch size.' + + num_patches = (image_height // patch_height) * (image_width // patch_width) + patch_dim = channels * patch_height * patch_width + assert pool in {'cls', 'mean'}, 'pool type must be either cls (cls token) or mean (mean pooling)' + + self.to_patch_embedding = nn.Sequential( + Rearrange('b c (h p1) (w p2) -> b (h w) (p1 p2 c)', p1=patch_height, p2=patch_width), + nn.LayerNorm(patch_dim), + nn.Linear(patch_dim, dim), + nn.LayerNorm(dim), + ) + + self.pos_embedding = nn.Parameter(torch.randn(1, num_patches + 1, dim)) + self.cls_token = nn.Parameter(torch.randn(1, 1, dim)) + self.dropout = nn.Dropout(emb_dropout) + + self.transformer = Transformer(dim, depth, heads, dim_head, mlp_dim, dropout) + + self.pool = pool + self.to_latent = nn.Identity() + + self.mlp_head = nn.Linear(dim, num_classes) + + def forward(self, img): + x = self.to_patch_embedding(img) + b, n, _ = x.shape + + cls_tokens = repeat(self.cls_token, '1 1 d -> b 1 d', b=b) + x = torch.cat((cls_tokens, x), dim=1) + x += self.pos_embedding[:, :(n + 1)] + x = self.dropout(x) + + x = self.transformer(x) + + x = x.mean(dim=1) if self.pool == 'mean' else x[:, 0] + + x = self.to_latent(x) + return self.mlp_head(x) diff --git a/contrast/.idea/.gitignore b/contrast/resnet_vit/.idea/.gitignore similarity index 100% rename from contrast/.idea/.gitignore rename to contrast/resnet_vit/.idea/.gitignore diff --git a/contrast/.idea/contrastInference.iml b/contrast/resnet_vit/.idea/contrastInference.iml similarity index 100% rename from contrast/.idea/contrastInference.iml rename to contrast/resnet_vit/.idea/contrastInference.iml diff --git a/contrast/.idea/deployment.xml b/contrast/resnet_vit/.idea/deployment.xml similarity index 100% rename from contrast/.idea/deployment.xml rename to contrast/resnet_vit/.idea/deployment.xml diff --git a/contrast/.idea/inspectionProfiles/Project_Default.xml b/contrast/resnet_vit/.idea/inspectionProfiles/Project_Default.xml similarity index 100% rename from contrast/.idea/inspectionProfiles/Project_Default.xml rename to contrast/resnet_vit/.idea/inspectionProfiles/Project_Default.xml diff --git a/contrast/.idea/inspectionProfiles/profiles_settings.xml b/contrast/resnet_vit/.idea/inspectionProfiles/profiles_settings.xml similarity index 100% rename from contrast/.idea/inspectionProfiles/profiles_settings.xml rename to contrast/resnet_vit/.idea/inspectionProfiles/profiles_settings.xml diff --git a/contrast/.idea/misc.xml b/contrast/resnet_vit/.idea/misc.xml similarity index 100% rename from contrast/.idea/misc.xml rename to contrast/resnet_vit/.idea/misc.xml diff --git a/contrast/.idea/modules.xml b/contrast/resnet_vit/.idea/modules.xml similarity index 100% rename from contrast/.idea/modules.xml rename to contrast/resnet_vit/.idea/modules.xml diff --git a/contrast/__init__.py b/contrast/resnet_vit/__init__.py similarity index 100% rename from contrast/__init__.py rename to contrast/resnet_vit/__init__.py diff --git a/contrast/__pycache__/__init__.cpython-39.pyc b/contrast/resnet_vit/__pycache__/__init__.cpython-39.pyc similarity index 100% rename from contrast/__pycache__/__init__.cpython-39.pyc rename to contrast/resnet_vit/__pycache__/__init__.cpython-39.pyc diff --git a/contrast/__pycache__/config.cpython-38.pyc b/contrast/resnet_vit/__pycache__/config.cpython-38.pyc similarity index 100% rename from contrast/__pycache__/config.cpython-38.pyc rename to contrast/resnet_vit/__pycache__/config.cpython-38.pyc diff --git a/contrast/resnet_vit/__pycache__/config.cpython-39.pyc b/contrast/resnet_vit/__pycache__/config.cpython-39.pyc new file mode 100644 index 0000000..ba23f1d Binary files /dev/null and b/contrast/resnet_vit/__pycache__/config.cpython-39.pyc differ diff --git a/contrast/__pycache__/inference.cpython-39.pyc b/contrast/resnet_vit/__pycache__/inference.cpython-39.pyc similarity index 100% rename from contrast/__pycache__/inference.cpython-39.pyc rename to contrast/resnet_vit/__pycache__/inference.cpython-39.pyc diff --git a/contrast/resnet_vit/config.py b/contrast/resnet_vit/config.py new file mode 100644 index 0000000..aea7680 --- /dev/null +++ b/contrast/resnet_vit/config.py @@ -0,0 +1,84 @@ +import torch +import torchvision.transforms as T + + +class Config: + # network settings + backbone = 'vit' # [resnet18, mobilevit_s, mobilenet_v2, mobilenetv3_small, mobilenetv3_large, mobilenet_v1, PPLCNET_x1_0, PPLCNET_x0_5, PPLCNET_x2_5] + metric = 'softmax' # [cosface, arcface, softmax] + cbam = True + embedding_size = 256 # 256 + drop_ratio = 0.5 + img_size = 224 + + teacher = 'vit' # [resnet18, mobilevit_s, mobilenet_v2, mobilenetv3_small, mobilenetv3_large, mobilenet_v1, PPLCNET_x1_0, PPLCNET_x0_5, PPLCNET_x2_5] + student = 'resnet' + # data preprocess + # input_shape = [1, 128, 128] + """transforms.RandomCrop(size), + transforms.RandomVerticalFlip(p=0.5), + transforms.RandomHorizontalFlip(), + RandomRotate(15, 0.3), + # RandomGaussianBlur()""" + + train_transform = T.Compose([ + T.ToTensor(), + T.Resize((img_size, img_size)), + # T.RandomCrop(img_size*4//5), + # T.RandomHorizontalFlip(p=0.5), + T.RandomRotation(180), + T.ColorJitter(brightness=0.5), + T.ConvertImageDtype(torch.float32), + T.Normalize(mean=[0.5], std=[0.5]), + ]) + test_transform = T.Compose([ + T.ToTensor(), + T.Resize((img_size, img_size)), + T.ConvertImageDtype(torch.float32), + T.Normalize(mean=[0.5], std=[0.5]), + ]) + + # dataset + train_root = './data/2250_train/train' # 初始筛选过一次的数据集 + # train_root = './data/0625_train/train' + test_root = "./data/2250_train/val/" + # test_root = "./data/0625_train/val" + + test_list = "./data/2250_train/val_pair.txt" + test_group_json = "./data/2250_train/cross_same.json" + # test_group_json = "./data/0625_train/cross_same.json" + # test_list = "./data/test_data_100/val_pair.txt" + + # training settings + checkpoints = "checkpoints/vit_b_16_0815/" # [resnet18, mobilevit_s, mobilenet_v2, mobilenetv3] + restore = True + # restore_model = "checkpoints/renet18_2250_0315/best_resnet18_2250_0315.pth" # best_resnet18_1491_0306.pth + restore_model = "checkpoints/vit_b_16_0730/best.pth" # best_resnet18_1491_0306.pth + + # test_model = "./checkpoints/renet18_1887_0311/best_resnet18_1887_0311.pth" + testbackbone = 'resnet18' # [resnet18, mobilevit_s, mobilenet_v2, mobilenetv3_small, mobilenetv3_large, mobilenet_v1, PPLCNET_x1_0, PPLCNET_x0_5] + # test_val = "./data/2250_train" + test_val = "./data/0625_train" + test_model = "checkpoints/resnet18_0721/best.pth" + + train_batch_size = 128 # 256 + test_batch_size = 256 # 256 + + + epoch = 300 + optimizer = 'adamw' # ['sgd', 'adam'， 'adamw'] + lr = 1e-3 # 1e-2 + lr_step = 10 # 10 + lr_decay = 0.95 # 0.98 + weight_decay = 5e-4 + loss = 'focal_loss' # ['focal_loss', 'cross_entropy'] + device = torch.device('cuda:1' if torch.cuda.is_available() else 'cpu') + # device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu') + + pin_memory = True # if memory is large, set it True to speed up a bit + num_workers = 4 # dataloader + + group_test = True + # group_test = False + +config = Config() \ No newline at end of file diff --git a/contrast/inference.py b/contrast/resnet_vit/inference.py similarity index 100% rename from contrast/inference.py rename to contrast/resnet_vit/inference.py diff --git a/contrast/resnet_vit/model/__init__.py b/contrast/resnet_vit/model/__init__.py new file mode 100644 index 0000000..9eebc77 --- /dev/null +++ b/contrast/resnet_vit/model/__init__.py @@ -0,0 +1 @@ +from .resnet_pre import resnet18, resnet34, resnet50, resnet14 diff --git a/contrast/resnet_vit/model/__pycache__/__init__.cpython-38.pyc b/contrast/resnet_vit/model/__pycache__/__init__.cpython-38.pyc new file mode 100644 index 0000000..160bc37 Binary files /dev/null and b/contrast/resnet_vit/model/__pycache__/__init__.cpython-38.pyc differ diff --git a/contrast/resnet_vit/model/__pycache__/__init__.cpython-39.pyc b/contrast/resnet_vit/model/__pycache__/__init__.cpython-39.pyc new file mode 100644 index 0000000..539ec67 Binary files /dev/null and b/contrast/resnet_vit/model/__pycache__/__init__.cpython-39.pyc differ diff --git a/contrast/resnet_vit/model/__pycache__/resnet_pre.cpython-38.pyc b/contrast/resnet_vit/model/__pycache__/resnet_pre.cpython-38.pyc new file mode 100644 index 0000000..199085d Binary files /dev/null and b/contrast/resnet_vit/model/__pycache__/resnet_pre.cpython-38.pyc differ diff --git a/contrast/resnet_vit/model/__pycache__/resnet_pre.cpython-39.pyc b/contrast/resnet_vit/model/__pycache__/resnet_pre.cpython-39.pyc new file mode 100644 index 0000000..0301f55 Binary files /dev/null and b/contrast/resnet_vit/model/__pycache__/resnet_pre.cpython-39.pyc differ diff --git a/contrast/resnet_vit/model/resnet_pre.py b/contrast/resnet_vit/model/resnet_pre.py new file mode 100644 index 0000000..5e52ad9 --- /dev/null +++ b/contrast/resnet_vit/model/resnet_pre.py @@ -0,0 +1,462 @@ +import torch +import torch.nn as nn +from config import config as conf + +try: + from torch.hub import load_state_dict_from_url +except ImportError: + from torch.utils.model_zoo import load_url as load_state_dict_from_url +# from .utils import load_state_dict_from_url + +__all__ = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101', + 'resnet152', 'resnext50_32x4d', 'resnext101_32x8d', + 'wide_resnet50_2', 'wide_resnet101_2'] + +model_urls = { + 'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth', + 'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth', + 'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth', + 'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth', + 'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth', + 'resnext50_32x4d': 'https://download.pytorch.org/models/resnext50_32x4d-7cdf4587.pth', + 'resnext101_32x8d': 'https://download.pytorch.org/models/resnext101_32x8d-8ba56ff5.pth', + 'wide_resnet50_2': 'https://download.pytorch.org/models/wide_resnet50_2-95faca4d.pth', + 'wide_resnet101_2': 'https://download.pytorch.org/models/wide_resnet101_2-32ee1156.pth', +} + + +def conv3x3(in_planes, out_planes, stride=1, groups=1, dilation=1): + """3x3 convolution with padding""" + return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride, + padding=dilation, groups=groups, bias=False, dilation=dilation) + +def conv1x1(in_planes, out_planes, stride=1): + """1x1 convolution""" + return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False) + + +class SpatialAttention(nn.Module): + def __init__(self, kernel_size=7): + super(SpatialAttention, self).__init__() + + assert kernel_size in (3, 7), 'kernel size must be 3 or 7' + padding = 3 if kernel_size == 7 else 1 + + self.conv1 = nn.Conv2d(2, 1, kernel_size, padding=padding, bias=False) + self.sigmoid = nn.Sigmoid() + + def forward(self, x): + avg_out = torch.mean(x, dim=1, keepdim=True) + max_out, _ = torch.max(x, dim=1, keepdim=True) + x = torch.cat([avg_out, max_out], dim=1) + x = self.conv1(x) + return self.sigmoid(x) + +class BasicBlock(nn.Module): + expansion = 1 + + def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1, + base_width=64, dilation=1, norm_layer=None, cam=False, bam=False): + super(BasicBlock, self).__init__() + if norm_layer is None: + norm_layer = nn.BatchNorm2d + if groups != 1 or base_width != 64: + raise ValueError('BasicBlock only supports groups=1 and base_width=64') + if dilation > 1: + raise NotImplementedError("Dilation > 1 not supported in BasicBlock") + self.cam = cam + self.bam = bam + # Both self.conv1 and self.downsample layers downsample the input when stride != 1 + self.conv1 = conv3x3(inplanes, planes, stride) + self.bn1 = norm_layer(planes) + self.relu = nn.ReLU(inplace=True) + self.conv2 = conv3x3(planes, planes) + self.bn2 = norm_layer(planes) + self.downsample = downsample + self.stride = stride + if self.cam: + if planes == 64: + self.globalAvgPool = nn.AvgPool2d(56, stride=1) + elif planes == 128: + self.globalAvgPool = nn.AvgPool2d(28, stride=1) + elif planes == 256: + self.globalAvgPool = nn.AvgPool2d(14, stride=1) + elif planes == 512: + self.globalAvgPool = nn.AvgPool2d(7, stride=1) + + self.fc1 = nn.Linear(in_features=planes, out_features=round(planes / 16)) + self.fc2 = nn.Linear(in_features=round(planes / 16), out_features=planes) + self.sigmod = nn.Sigmoid() + if self.bam: + self.bam = SpatialAttention() + + def forward(self, x): + identity = x + + out = self.conv1(x) + out = self.bn1(out) + out = self.relu(out) + + out = self.conv2(out) + out = self.bn2(out) + + if self.downsample is not None: + identity = self.downsample(x) + + if self.cam: + ori_out = self.globalAvgPool(out) + out = out.view(out.size(0), -1) + out = self.fc1(out) + out = self.relu(out) + out = self.fc2(out) + out = self.sigmod(out) + out = out.view(out.size(0), out.size(-1), 1, 1) + out = out * ori_out + + if self.bam: + out = out*self.bam(out) + + out += identity + out = self.relu(out) + + return out + + +class Bottleneck(nn.Module): + # Bottleneck in torchvision places the stride for downsampling at 3x3 convolution(self.conv2) + # while original implementation places the stride at the first 1x1 convolution(self.conv1) + # according to "Deep residual learning for image recognition"https://arxiv.org/abs/1512.03385. + # This variant is also known as ResNet V1.5 and improves accuracy according to + # https://ngc.nvidia.com/catalog/model-scripts/nvidia:resnet_50_v1_5_for_pytorch. + + expansion = 4 + + def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1, + base_width=64, dilation=1, norm_layer=None, cam=False, bam=False): + super(Bottleneck, self).__init__() + if norm_layer is None: + norm_layer = nn.BatchNorm2d + width = int(planes * (base_width / 64.)) * groups + self.cam = cam + self.bam = bam + # Both self.conv2 and self.downsample layers downsample the input when stride != 1 + self.conv1 = conv1x1(inplanes, width) + self.bn1 = norm_layer(width) + self.conv2 = conv3x3(width, width, stride, groups, dilation) + self.bn2 = norm_layer(width) + self.conv3 = conv1x1(width, planes * self.expansion) + self.bn3 = norm_layer(planes * self.expansion) + self.relu = nn.ReLU(inplace=True) + self.downsample = downsample + self.stride = stride + if self.cam: + if planes == 64: + self.globalAvgPool = nn.AvgPool2d(56, stride=1) + elif planes == 128: + self.globalAvgPool = nn.AvgPool2d(28, stride=1) + elif planes == 256: + self.globalAvgPool = nn.AvgPool2d(14, stride=1) + elif planes == 512: + self.globalAvgPool = nn.AvgPool2d(7, stride=1) + + self.fc1 = nn.Linear(planes * self.expansion, round(planes / 4)) + self.fc2 = nn.Linear(round(planes / 4), planes * self.expansion) + self.sigmod = nn.Sigmoid() + + def forward(self, x): + identity = x + + out = self.conv1(x) + out = self.bn1(out) + out = self.relu(out) + + out = self.conv2(out) + out = self.bn2(out) + out = self.relu(out) + + out = self.conv3(out) + out = self.bn3(out) + + if self.downsample is not None: + identity = self.downsample(x) + + if self.cam: + ori_out = self.globalAvgPool(out) + out = out.view(out.size(0), -1) + out = self.fc1(out) + out = self.relu(out) + out = self.fc2(out) + out = self.sigmod(out) + out = out.view(out.size(0), out.size(-1), 1, 1) + out = out * ori_out + out += identity + out = self.relu(out) + return out + + +class ResNet(nn.Module): + + def __init__(self, block, layers, num_classes=conf.embedding_size, zero_init_residual=False, + groups=1, width_per_group=64, replace_stride_with_dilation=None, + norm_layer=None, scale=0.75): + super(ResNet, self).__init__() + if norm_layer is None: + norm_layer = nn.BatchNorm2d + self._norm_layer = norm_layer + + self.inplanes = 64 + self.dilation = 1 + if replace_stride_with_dilation is None: + # each element in the tuple indicates if we should replace + # the 2x2 stride with a dilated convolution instead + replace_stride_with_dilation = [False, False, False] + if len(replace_stride_with_dilation) != 3: + raise ValueError("replace_stride_with_dilation should be None " + "or a 3-element tuple, got {}".format(replace_stride_with_dilation)) + self.groups = groups + self.base_width = width_per_group + self.conv1 = nn.Conv2d(3, self.inplanes, kernel_size=7, stride=2, padding=3, + bias=False) + self.bn1 = norm_layer(self.inplanes) + self.relu = nn.ReLU(inplace=True) + self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) + self.layer1 = self._make_layer(block, int(64*scale), layers[0]) + self.layer2 = self._make_layer(block, int(128*scale), layers[1], stride=2, + dilate=replace_stride_with_dilation[0]) + self.layer3 = self._make_layer(block, int(256*scale), layers[2], stride=2, + dilate=replace_stride_with_dilation[1]) + self.layer4 = self._make_layer(block, int(512*scale), layers[3], stride=2, + dilate=replace_stride_with_dilation[2]) + self.avgpool = nn.AdaptiveAvgPool2d((1, 1)) + self.fc = nn.Linear(int(512 * block.expansion*scale), num_classes) + + for m in self.modules(): + if isinstance(m, nn.Conv2d): + nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') + elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)): + nn.init.constant_(m.weight, 1) + nn.init.constant_(m.bias, 0) + + # Zero-initialize the last BN in each residual branch, + # so that the residual branch starts with zeros, and each residual block behaves like an identity. + # This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677 + if zero_init_residual: + for m in self.modules(): + if isinstance(m, Bottleneck): + nn.init.constant_(m.bn3.weight, 0) + elif isinstance(m, BasicBlock): + nn.init.constant_(m.bn2.weight, 0) + + def _make_layer(self, block, planes, blocks, stride=1, dilate=False): + norm_layer = self._norm_layer + downsample = None + previous_dilation = self.dilation + if dilate: + self.dilation *= stride + stride = 1 + if stride != 1 or self.inplanes != planes * block.expansion: + downsample = nn.Sequential( + conv1x1(self.inplanes, planes * block.expansion, stride), + norm_layer(planes * block.expansion), + ) + + layers = [] + layers.append(block(self.inplanes, planes, stride, downsample, self.groups, + self.base_width, previous_dilation, norm_layer)) + self.inplanes = planes * block.expansion + for _ in range(1, blocks): + layers.append(block(self.inplanes, planes, groups=self.groups, + base_width=self.base_width, dilation=self.dilation, + norm_layer=norm_layer)) + return nn.Sequential(*layers) + + def _forward_impl(self, x): + # See note [TorchScript super()] + x = self.conv1(x) + x = self.bn1(x) + x = self.relu(x) + x = self.maxpool(x) + + x = self.layer1(x) + x = self.layer2(x) + x = self.layer3(x) + x = self.layer4(x) + + # print('poolBefore', x.shape) + x = self.avgpool(x) + # print('poolAfter', x.shape) + x = torch.flatten(x, 1) + # print('fcBefore',x.shape) + x = self.fc(x) + + # print('fcAfter',x.shape) + + return x + + def forward(self, x): + return self._forward_impl(x) + + +# def _resnet(arch, block, layers, pretrained, progress, **kwargs): +# model = ResNet(block, layers, **kwargs) +# if pretrained: +# state_dict = load_state_dict_from_url(model_urls[arch], +# progress=progress) +# model.load_state_dict(state_dict, strict=False) +# return model +def _resnet(arch, block, layers, pretrained, progress, **kwargs): + model = ResNet(block, layers, **kwargs) + if pretrained: + state_dict = load_state_dict_from_url(model_urls[arch], + progress=progress) + + src_state_dict = state_dict + target_state_dict = model.state_dict() + skip_keys = [] + # skip mismatch size tensors in case of pretraining + for k in src_state_dict.keys(): + if k not in target_state_dict: + continue + if src_state_dict[k].size() != target_state_dict[k].size(): + skip_keys.append(k) + for k in skip_keys: + del src_state_dict[k] + missing_keys, unexpected_keys = model.load_state_dict(src_state_dict, strict=False) + + return model + + +def resnet14(pretrained=True, progress=True, **kwargs): + r"""ResNet-14 model from + `"Deep Residual Learning for Image Recognition" `_ + + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + progress (bool): If True, displays a progress bar of the download to stderr + """ + return _resnet('resnet18', BasicBlock, [2, 1, 1, 2], pretrained, progress, + **kwargs) + + +def resnet18(pretrained=True, progress=True, **kwargs): + r"""ResNet-18 model from + `"Deep Residual Learning for Image Recognition" `_ + + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + progress (bool): If True, displays a progress bar of the download to stderr + """ + return _resnet('resnet18', BasicBlock, [2, 2, 2, 2], pretrained, progress, + **kwargs) + + +def resnet34(pretrained=False, progress=True, **kwargs): + r"""ResNet-34 model from + `"Deep Residual Learning for Image Recognition" `_ + + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + progress (bool): If True, displays a progress bar of the download to stderr + """ + return _resnet('resnet34', BasicBlock, [3, 4, 6, 3], pretrained, progress, + **kwargs) + + +def resnet50(pretrained=False, progress=True, **kwargs): + r"""ResNet-50 model from + `"Deep Residual Learning for Image Recognition" `_ + + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + progress (bool): If True, displays a progress bar of the download to stderr + """ + return _resnet('resnet50', Bottleneck, [3, 4, 6, 3], pretrained, progress, + **kwargs) + + +def resnet101(pretrained=False, progress=True, **kwargs): + r"""ResNet-101 model from + `"Deep Residual Learning for Image Recognition" `_ + + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + progress (bool): If True, displays a progress bar of the download to stderr + """ + return _resnet('resnet101', Bottleneck, [3, 4, 23, 3], pretrained, progress, + **kwargs) + + +def resnet152(pretrained=False, progress=True, **kwargs): + r"""ResNet-152 model from + `"Deep Residual Learning for Image Recognition" `_ + + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + progress (bool): If True, displays a progress bar of the download to stderr + """ + return _resnet('resnet152', Bottleneck, [3, 8, 36, 3], pretrained, progress, + **kwargs) + + +def resnext50_32x4d(pretrained=False, progress=True, **kwargs): + r"""ResNeXt-50 32x4d model from + `"Aggregated Residual Transformation for Deep Neural Networks" `_ + + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + progress (bool): If True, displays a progress bar of the download to stderr + """ + kwargs['groups'] = 32 + kwargs['width_per_group'] = 4 + return _resnet('resnext50_32x4d', Bottleneck, [3, 4, 6, 3], + pretrained, progress, **kwargs) + + +def resnext101_32x8d(pretrained=False, progress=True, **kwargs): + r"""ResNeXt-101 32x8d model from + `"Aggregated Residual Transformation for Deep Neural Networks" `_ + + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + progress (bool): If True, displays a progress bar of the download to stderr + """ + kwargs['groups'] = 32 + kwargs['width_per_group'] = 8 + return _resnet('resnext101_32x8d', Bottleneck, [3, 4, 23, 3], + pretrained, progress, **kwargs) + + +def wide_resnet50_2(pretrained=False, progress=True, **kwargs): + r"""Wide ResNet-50-2 model from + `"Wide Residual Networks" `_ + + The model is the same as ResNet except for the bottleneck number of channels + which is twice larger in every block. The number of channels in outer 1x1 + convolutions is the same, e.g. last block in ResNet-50 has 2048-512-2048 + channels, and in Wide ResNet-50-2 has 2048-1024-2048. + + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + progress (bool): If True, displays a progress bar of the download to stderr + """ + kwargs['width_per_group'] = 64 * 2 + return _resnet('wide_resnet50_2', Bottleneck, [3, 4, 6, 3], + pretrained, progress, **kwargs) + + +def wide_resnet101_2(pretrained=False, progress=True, **kwargs): + r"""Wide ResNet-101-2 model from + `"Wide Residual Networks" `_ + + The model is the same as ResNet except for the bottleneck number of channels + which is twice larger in every block. The number of channels in outer 1x1 + convolutions is the same, e.g. last block in ResNet-50 has 2048-512-2048 + channels, and in Wide ResNet-50-2 has 2048-1024-2048. + + Args: + pretrained (bool): If True, returns a model pre-trained on ImageNet + progress (bool): If True, displays a progress bar of the download to stderr + """ + kwargs['width_per_group'] = 64 * 2 + return _resnet('wide_resnet101_2', Bottleneck, [3, 4, 23, 3], + pretrained, progress, **kwargs) diff --git a/contrast/results.txt b/contrast/results.txt new file mode 100644 index 0000000..f7f6466 --- /dev/null +++ b/contrast/results.txt @@ -0,0 +1,124 @@ +same, 6954256905319, 20240715-144627_6954256905319, 0.211, 0.750, 0.630 +same, 6930364600025, 20240719-143603_6930364600025, 0.283, 0.771, 0.616 +same, 6924509902839, 20240719-143630_6924509902839, 0.182, 0.727, 0.462 +same, 6902007115524, 20240719-143729_6902007115524, 0.221, 0.723, 0.519 +same, 6930628801472, 20240719-143807_6930628801472, 0.249, 0.804, 0.672 +same, 6908355010450, 20240719-144335_6908355010450, 0.255, 0.812, 0.575 +same, 6924509902754, 20240719-144454_6924509902754, 0.458, 0.882, 0.836 +same, 6925843410158, 20240719-144916_6925843410158, 0.229, 0.829, 0.686 +same, 6941499112529, 20240719-145022_6941499112529, 0.190, 0.795, 0.359 +same, 6918149576536, 20240719-145058_6918149576536, 0.493, 0.906, 0.765 +same, 6928002374551, 20240719-145542_6928002374551, 0.165, 0.773, 0.531 +same, 6930364600001, 20240719-150000_6930364600001, 0.322, 0.831, 0.736 +same, 6937962111526, 20240719-150209_6937962111526, 0.225, 0.794, 0.605 +same, 6924509902839, 20240719-150345_6924509902839, 0.161, 0.816, 0.502 +same, 6924509902839, 20240719-150458_6924509902839, 0.319, 0.746, 0.623 +same, 6924187895423, 20240719-150623_6924187895423, 0.430, 0.906, 0.903 +same, 6916749306102, 20240719-150716_6916749306102, 0.269, 0.764, 0.737 +same, 6951469100067, 20240719-150750_6951469100067, 0.547, 0.938, 0.896 +same, 6958350501101, 20240719-151215_6958350501101, 0.578, 0.906, 0.877 +same, 6951469100067, 20240719-151325_6951469100067, 0.610, 0.930, 0.904 +same, 6937962111526, 20240719-151349_6937962111526, 0.138, 0.762, 0.496 +same, 6916749306102, 20240719-151416_6916749306102, 0.243, 0.728, 0.630 +same, 6930364600001, 20240719-151535_6930364600001, 0.144, 0.787, 0.482 +same, 6911988027191, 20240719-151724_6911988027191, 0.035, 0.502, 0.076 +same, 6972219890632, 20240719-151800_6972219890632, 0.315, 0.880, 0.700 +same, 6901668935908, 20240719-152125_6901668935908, 0.203, 0.856, 0.426 +same, 6930364600001, 20240719-153117_6930364600001, 0.057, 0.775, 0.165 +same, 6902265587712, 20240722-144630_6902265587712, 0.364, 0.833, 0.669 +same, 4897073991658, 20240722-160713_4897073991658, 0.101, 0.514, 0.247 +same, 4892655902237, 20240722-160857_4892655902237, 0.459, 0.760, 0.812 +same, 6920907803020, 20240722-161533_6920907803020, 0.211, 0.769, 0.678 +same, 6920912342002, 20240722-163428_6920912342002, 0.281, 0.825, 0.853 +same, 6916196422325, 20240722-164859_6916196422325, 0.392, 0.837, 0.665 +same, 6901668006011, 20240722-165548_6901668006011, 0.284, 0.752, 0.530 +same, 6937221716189, 20240723-094426_6937221716189, 0.407, 0.850, 0.705 +same, 4897098800133, 20240723-094447_4897098800133, 0.382, 0.846, 0.784 +same, 6970470331529, 20240723-094617_6970470331529, 0.238, 0.925, 0.586 +same, 6918598120014, 20240723-095103_6918598120014, 0.407, 0.905, 0.819 +same, 6938737200650, 20240723-095527_6938737200650, 0.208, 0.825, 0.543 +same, 6902890022558, 20240723-095858_6902890022558, 0.162, 0.769, 0.628 +same, 6925303795535, 20240723-100031_6925303795535, 0.304, 0.835, 0.671 +same, 6920382723257, 20240723-100113_6920382723257, 0.274, 0.854, 0.539 +same, 6938737200650, 20240723-100416_6938737200650, 0.237, 0.719, 0.581 +same, 6918598120014, 20240723-100501_6918598120014, 0.099, 0.791, 0.209 +same, 6937221716189, 20240723-100726_6937221716189, 0.347, 0.771, 0.611 +same, 4897098800133, 20240723-100845_4897098800133, 0.304, 0.770, 0.718 +same, 6918598120014, 20240723-101829_6918598120014, 0.502, 0.919, 0.857 +same, 6920382723257, 20240723-102033_6920382723257, 0.343, 0.849, 0.748 +same, 6902890022558, 20240723-102206_6902890022558, 0.117, 0.745, 0.389 +same, 6970470331529, 20240723-102409_6970470331529, 0.102, 0.734, 0.305 +same, 6938737200650, 20240723-102515_6938737200650, 0.183, 0.725, 0.469 +same, 6937221716189, 20240723-102833_6937221716189, 0.067, 0.623, 0.175 +same, 6937451883132, 20240723-110216_6937451883132, 0.323, 0.860, 0.717 +same, 6973337010216, 20240723-110224_6973337010216, 0.202, 0.760, 0.622 +same, 6937451883132, 20240723-111446_6937451883132, 0.367, 0.817, 0.806 +same, 6973337010216, 20240723-111521_6973337010216, 0.140, 0.777, 0.414 +same, 6924187828544, 20240723-111602_6924187828544, 0.359, 0.859, 0.718 +same, 6934024590268, 20240723-112004_6934024590268, 0.184, 0.727, 0.583 +same, 6924187828544, 20240723-112505_6924187828544, 0.281, 0.875, 0.573 +same, 6974158894528, 20240723-155423_6974158894528, 0.433, 0.890, 0.788 +same, 6970399922518, 20240723-160227_6970399922518, 0.078, 0.564, 0.215 +same, 6974158894528, 20240723-160942_6974158894528, 0.417, 0.895, 0.832 +diff, 6902007115524, 20240715-144627_6954256905319, -0.042, 0.389, -0.098 +diff, 6924509902839, 20240719-143603_6930364600025, 0.028, 0.527, 0.056 +diff, 6934024590268, 20240719-143630_6924509902839, -0.003, 0.922, -0.011 +diff, 6918598120014, 20240719-143729_6902007115524, 0.061, 0.422, 0.138 +diff, 6916196422325, 20240719-143807_6930628801472, 0.046, 0.412, 0.077 +diff, 6973337010216, 20240719-144335_6908355010450, 0.019, 0.578, 0.058 +diff, 6902265587712, 20240719-144454_6924509902754, -0.026, 0.471, -0.042 +diff, 4892655902237, 20240719-144916_6925843410158, -0.066, 0.312, -0.140 +diff, 6958350501101, 20240719-145022_6941499112529, 0.028, 0.355, 0.043 +diff, 6925843410158, 20240719-145058_6918149576536, -0.026, 0.432, -0.064 +diff, 4897098800133, 20240719-145542_6928002374551, 0.017, 0.265, 0.044 +diff, 6930628801472, 20240719-150000_6930364600001, 0.037, 0.559, 0.095 +diff, 6974158894528, 20240719-150209_6937962111526, -0.064, 0.433, -0.137 +diff, 6930628801472, 20240719-150345_6924509902839, 0.001, 0.509, 0.004 +diff, 6937221716189, 20240719-150458_6924509902839, -0.038, 0.455, -0.071 +diff, 6902007115524, 20240719-150623_6924187895423, 0.020, 0.429, 0.046 +diff, 6930364600001, 20240719-150716_6916749306102, 0.051, 0.454, 0.138 +diff, 6938737200650, 20240719-150750_6951469100067, -0.007, 0.403, -0.017 +diff, 6924187895423, 20240719-151215_6958350501101, 0.012, 0.411, 0.021 +diff, 6974158894528, 20240719-151325_6951469100067, -0.002, 0.394, -0.004 +diff, 6916196422325, 20240719-151349_6937962111526, -0.013, 0.322, -0.026 +diff, 6920907803020, 20240719-151416_6916749306102, 0.012, 0.501, 0.034 +diff, 6941499112529, 20240719-151535_6930364600001, 0.003, 0.467, 0.007 +diff, 6902265587712, 20240719-151724_6911988027191, -0.001, 0.464, -0.002 +diff, 6916196422325, 20240719-151800_6972219890632, -0.033, 0.330, -0.058 +diff, 6930364600025, 20240719-152125_6901668935908, -0.073, 0.411, -0.174 +diff, 6916749306102, 20240719-153117_6930364600001, 0.082, 0.604, 0.245 +diff, 6902890022558, 20240722-144630_6902265587712, -0.000, 0.477, -0.001 +diff, 6920907803020, 20240722-160713_4897073991658, 0.083, 0.569, 0.255 +diff, 6937962111526, 20240722-160857_4892655902237, -0.026, 0.351, -0.073 +diff, 6925303795535, 20240722-161533_6920907803020, 0.149, 0.645, 0.421 +diff, 6934024590268, 20240722-163428_6920912342002, 0.062, 0.519, 0.215 +diff, 6972219890632, 20240722-164859_6916196422325, 0.048, 0.512, 0.102 +diff, 6972219890632, 20240722-165548_6901668006011, 0.035, 0.369, 0.067 +diff, 6918598120014, 20240723-094426_6937221716189, 0.078, 0.435, 0.142 +diff, 6937451883132, 20240723-094447_4897098800133, 0.071, 0.606, 0.157 +diff, 6954256905319, 20240723-094617_6970470331529, 0.019, 0.570, 0.062 +diff, 6930628801472, 20240723-095103_6918598120014, 0.022, 0.404, 0.057 +diff, 6902007115524, 20240723-095527_6938737200650, 0.013, 0.421, 0.030 +diff, 6930364600001, 20240723-095858_6902890022558, -0.014, 0.410, -0.039 +diff, 6930628801472, 20240723-100031_6925303795535, 0.005, 0.471, 0.013 +diff, 6930628801472, 20240723-100113_6920382723257, -0.009, 0.458, -0.023 +diff, 6924187828544, 20240723-100416_6938737200650, 0.127, 0.534, 0.266 +diff, 6920907803020, 20240723-100501_6918598120014, 0.024, 0.569, 0.062 +diff, 6918598120014, 20240723-100726_6937221716189, 0.038, 0.409, 0.070 +diff, 6930628801472, 20240723-100845_4897098800133, -0.000, 0.414, -0.000 +diff, 6902007115524, 20240723-101829_6918598120014, 0.027, 0.533, 0.047 +diff, 6970470331529, 20240723-102033_6920382723257, -0.001, 0.592, -0.003 +diff, 6974158894528, 20240723-102206_6902890022558, -0.039, 0.424, -0.089 +diff, 6920907803020, 20240723-102409_6970470331529, 0.046, 0.441, 0.164 +diff, 6970470331529, 20240723-102515_6938737200650, -0.048, 0.388, -0.107 +diff, 6924187895423, 20240723-102833_6937221716189, 0.066, 0.511, 0.169 +diff, 6920907803020, 20240723-110216_6937451883132, -0.051, 0.370, -0.128 +diff, 6924187828544, 20240723-110224_6973337010216, 0.033, 0.452, 0.078 +diff, 6908355010450, 20240723-111446_6937451883132, 0.137, 0.453, 0.264 +diff, 6930628801472, 20240723-111521_6973337010216, -0.001, 0.403, -0.003 +diff, 4897073991658, 20240723-111602_6924187828544, 0.032, 0.306, 0.062 +diff, 6937221716189, 20240723-112004_6934024590268, 0.056, 0.468, 0.125 +diff, 4897098800133, 20240723-112505_6924187828544, -0.039, 0.258, -0.080 +diff, 6916749306102, 20240723-155423_6974158894528, -0.001, 0.378, -0.003 +diff, 6973337010216, 20240723-160227_6970399922518, -0.015, 0.477, -0.048 +diff, 6958350501101, 20240723-160942_6974158894528, 0.042, 0.426, 0.062 diff --git a/contrast/test_ori.py b/contrast/test_ori.py new file mode 100644 index 0000000..7333a96 --- /dev/null +++ b/contrast/test_ori.py @@ -0,0 +1,476 @@ +# -*- coding: utf-8 -*- +import os +import os.path as osp +import pdb +import numpy as np +import shutil +from scipy.spatial.distance import cdist +import torch +import torch.nn as nn +from PIL import Image +import json +from config import config as conf +from model import resnet18 + +# from model import (mobilevit_s, resnet14, resnet18, resnet34, resnet50, mobilenet_v2, +# MobileNetV3_Small, mobilenet_v1, PPLCNET_x1_0, PPLCNET_x0_5, PPLCNET_x2_5) + + +import matplotlib.pyplot as plt +import statistics +embedding_size = conf.embedding_size +img_size = conf.img_size +device = conf.device + + +def unique_image(pair_list) -> set: + """Return unique image path in pair_list.txt""" + with open(pair_list, 'r') as fd: + pairs = fd.readlines() + unique = set() + for pair in pairs: + id1, id2, _ = pair.split() + unique.add(id1) + unique.add(id2) + return unique + + +def group_image(images: set, batch) -> list: + """Group image paths by batch size""" + images = list(images) + size = len(images) + res = [] + for i in range(0, size, batch): + end = min(batch + i, size) + res.append(images[i: end]) + return res + + +def _preprocess(images: list, transform) -> torch.Tensor: + res = [] + for img in images: + im = Image.open(img) + im = transform(im) + res.append(im) + # data = torch.cat(res, dim=0) # shape: (batch, 128, 128) + # data = data[:, None, :, :] # shape: (batch, 1, 128, 128) + data = torch.stack(res) + return data + + +def test_preprocess(images: list, transform) -> torch.Tensor: + res = [] + for img in images: + im = Image.open(img) + im = transform(im) + res.append(im) + # data = torch.cat(res, dim=0) # shape: (batch, 128, 128) + # data = data[:, None, :, :] # shape: (batch, 1, 128, 128) + data = torch.stack(res) + return data + + +def featurize(images: list, transform, net, device, train=False) -> dict: + """featurize each image and save into a dictionary + Args: + images: image paths + transform: test transform + net: pretrained model + device: cpu or cuda + Returns: + Dict (key: imagePath, value: feature) + """ + if train: + data = _preprocess(images, transform) + data = data.to(device) + net = net.to(device) + with torch.no_grad(): + features = net(data) + res = {img: feature for (img, feature) in zip(images, features)} + else: + data = test_preprocess(images, transform) + data = data.to(device) + net = net.to(device) + with torch.no_grad(): + features = net(data) + res = {img: feature for (img, feature) in zip(images, features)} + return res + +def inference_image(images: list, transform, net, device, bs=16, embedding_size=256) -> dict: + batch_patches = [] + patches = [] + for d, img in enumerate(images): + img = Image.open(img) + patch = transform(img) + + if str(device) != "cpu": + patch = patch.to(device).half() + else: + patch = patch.to(device) + + patches.append(patch) + if (d + 1) % bs == 0: + patches = torch.stack(patches, dim=0) + batch_patches.append(patches) + patches = [] + + if len(patches): + patches = torch.stack(patches, dim=0) + batch_patches.append(patches) + + features = np.zeros((0, embedding_size)) + for patches in batch_patches: + pred = net(patches) + pred[torch.isinf(pred)] = 1.0 + feat = pred.cpu().data.numpy() + features = np.vstack((features, feat)) + + + + return features + + + +def featurize_1(images: list, transform, net, device, train=False) -> dict: + """featurize each image and save into a dictionary + Args: + images: image paths + transform: test transform + net: pretrained model + device: cpu or cuda + Returns: + Dict (key: imagePath, value: feature) + """ + + data = test_preprocess(images, transform) + data = data.to(device) + net = net.to(device) + with torch.no_grad(): + features = net(data).data.numpy() + + return features + + + + +def cosin_metric(x1, x2): + return np.dot(x1, x2) / (np.linalg.norm(x1) * np.linalg.norm(x2)) + + +def threshold_search(y_score, y_true): + y_score = np.asarray(y_score) + y_true = np.asarray(y_true) + best_acc = 0 + best_th = 0 + for i in range(len(y_score)): + th = y_score[i] + y_test = (y_score >= th) + acc = np.mean((y_test == y_true).astype(int)) + if acc > best_acc: + best_acc = acc + best_th = th + return best_acc, best_th + + +def showgrid(recall, recall_TN, PrecisePos, PreciseNeg): + x = np.linspace(start=-1.0, stop=1.0, num=50, endpoint=True).tolist() + plt.figure(figsize=(10, 6)) + plt.plot(x, recall, color='red', label='recall') + plt.plot(x, recall_TN, color='black', label='recall_TN') + plt.plot(x, PrecisePos, color='blue', label='PrecisePos') + plt.plot(x, PreciseNeg, color='green', label='PreciseNeg') + plt.legend() + plt.xlabel('threshold') + # plt.ylabel('Similarity') + plt.grid(True, linestyle='--', alpha=0.5) + plt.savefig('accuracy_recall_grid.png') + plt.show() + plt.close() + + +def compute_accuracy_recall(score, labels): + th = 0.1 + squence = np.linspace(-1, 1, num=50) + # squence = [0.4] + recall, PrecisePos, PreciseNeg, recall_TN = [], [], [], [] + for th in squence: + t_score = (score > th) + t_labels = (labels == 1) + # print(t_score) + # print(t_labels) + TP = np.sum(np.logical_and(t_score, t_labels)) + FN = np.sum(np.logical_and(np.logical_not(t_score), t_labels)) + f_score = (score < th) + f_labels = (labels == 0) + TN = np.sum(np.logical_and(f_score, f_labels)) + FP = np.sum(np.logical_and(np.logical_not(f_score), f_labels)) + print("Threshold:{} TP:{},FP:{},TN:{},FN:{}".format(th, TP, FP, TN, FN)) + + PrecisePos.append(0 if TP / (TP + FP) == 'nan' else TP / (TP + FP)) + PreciseNeg.append(0 if TN == 0 else TN / (TN + FN)) + recall.append(0 if TP == 0 else TP / (TP + FN)) + recall_TN.append(0 if TN == 0 else TN / (TN + FP)) + showgrid(recall, recall_TN, PrecisePos, PreciseNeg) + + +def compute_accuracy(feature_dict, pair_list, test_root): + with open(pair_list, 'r') as f: + pairs = f.readlines() + + similarities = [] + labels = [] + for pair in pairs: + img1, img2, label = pair.split() + img1 = osp.join(test_root, img1) + img2 = osp.join(test_root, img2) + feature1 = feature_dict[img1].cpu().numpy() + feature2 = feature_dict[img2].cpu().numpy() + label = int(label) + + similarity = cosin_metric(feature1, feature2) + similarities.append(similarity) + labels.append(label) + + accuracy, threshold = threshold_search(similarities, labels) + # print('similarities >> {}'.format(similarities)) + # print('labels >> {}'.format(labels)) + compute_accuracy_recall(np.array(similarities), np.array(labels)) + return accuracy, threshold + + +def deal_group_pair(pairList1, pairList2): + allsimilarity = [] + one_similarity = [] + for pair1 in pairList1: + for pair2 in pairList2: + similarity = cosin_metric(pair1.cpu().numpy(), pair2.cpu().numpy()) + one_similarity.append(similarity) + allsimilarity.append(max(one_similarity)) # 最大值 + # allsimilarity.append(sum(one_similarity)/len(one_similarity)) # 均值 + # allsimilarity.append(statistics.median(one_similarity)) # 中位数 + # print(allsimilarity) + # print(labels) + return allsimilarity + +def compute_group_accuracy(content_list_read): + allSimilarity, allLabel= [], [] + for data_loaded in content_list_read: + one_group_list = [] + for i in range(2): + images = [osp.join(conf.test_val, img) for img in data_loaded[i]] + group = group_image(images, conf.test_batch_size) + d = featurize(group[0], conf.test_transform, model, conf.device) + one_group_list.append(d.values()) + similarity = deal_group_pair(one_group_list[0], one_group_list[1]) + allLabel.append(data_loaded[-1]) + allSimilarity.extend(similarity) + # print(allSimilarity) + # print(allLabel) + return allSimilarity, allLabel + +def compute_contrast_accuracy(content_list_read): + + npairs = 50 + + same_folder_pairs = content_list_read['same_folder_pairs'] + cross_folder_pairs = content_list_read['cross_folder_pairs'] + + npairs = min((len(same_folder_pairs), len(cross_folder_pairs))) + + + same_pairs = same_folder_pairs[:npairs] + cross_pairs = cross_folder_pairs[:npairs] + + same_pairs_similarity = [] + for i in range(len(same_pairs)): + images_a = [osp.join(conf.test_val, img) for img in same_pairs[i][0]] + images_b = [osp.join(conf.test_val, img) for img in same_pairs[i][1]] + + feats_a = inference_image(images_a, conf.test_transform, model, conf.device) + feats_b = inference_image(images_b, conf.test_transform, model, conf.device) + # matrix = 1- np.maximum(0.0, cdist(feats_a, feats_b, 'cosine')) + matrix = 1 - cdist(feats_a, feats_b, 'cosine') + + feats_am = np.mean(feats_a, axis=0, keepdims=True) + feats_bm = np.mean(feats_b, axis=0, keepdims=True) + matrixm = 1- np.maximum(0.0, cdist(feats_am, feats_bm, 'cosine')) + + same_pairs_similarity.append(np.mean(matrix)) + + '''保存相同 Barcode 图像对''' + # foldi = os.path.join('./result/same', f'{i}') + # if os.path.exists(foldi): + # shutil.rmtree(foldi) + # os.makedirs(foldi) + # else: + # os.makedirs(foldi) + # for ipt in range(len(images_a)): + # source_path = images_a[ipt] + # destination_path = os.path.join(foldi, f'a_{ipt}.png') + # shutil.copy2(source_path, destination_path) + # for ipt in range(len(images_b)): + # source_path = images_b[ipt] + # destination_path = os.path.join(foldi, f'b_{ipt}.png') + # shutil.copy2(source_path, destination_path) + + cross_pairs_similarity = [] + for i in range(len(cross_pairs)): + images_a = [osp.join(conf.test_val, img) for img in cross_pairs[i][0]] + images_b = [osp.join(conf.test_val, img) for img in cross_pairs[i][1]] + + feats_a = inference_image(images_a, conf.test_transform, model, conf.device) + feats_b = inference_image(images_b, conf.test_transform, model, conf.device) + # matrix = 1- np.maximum(0.0, cdist(feats_a, feats_b, 'cosine')) + matrix = 1 - cdist(feats_a, feats_b, 'cosine') + + feats_am = np.mean(feats_a, axis=0, keepdims=True) + feats_bm = np.mean(feats_b, axis=0, keepdims=True) + matrixm = 1- np.maximum(0.0, cdist(feats_am, feats_bm, 'cosine')) + + cross_pairs_similarity.append(np.mean(matrix)) + + '''保存不同 Barcode 图像对''' + # foldi = os.path.join('./result/cross', f'{i}') + # if os.path.exists(foldi): + # shutil.rmtree(foldi) + # os.makedirs(foldi) + # else: + # os.makedirs(foldi) + # for ipt in range(len(images_a)): + # source_path = images_a[ipt] + # destination_path = os.path.join(foldi, f'a_{ipt}.png') + # shutil.copy2(source_path, destination_path) + # for ipt in range(len(images_b)): + # source_path = images_b[ipt] + # destination_path = os.path.join(foldi, f'b_{ipt}.png') + # shutil.copy2(source_path, destination_path) + + + Thresh = np.linspace(-0.2, 1, 100) + + Same = np.array(same_pairs_similarity) + Cross = np.array(cross_pairs_similarity) + + fig, axs = plt.subplots(2, 1) + axs[0].hist(Same, bins=60, edgecolor='black') + axs[0].set_xlim([-0.2, 1]) + axs[0].set_title('Same Barcode') + + axs[1].hist(Cross, bins=60, edgecolor='black') + axs[1].set_xlim([-0.2, 1]) + axs[1].set_title('Cross Barcode') + + TPFN = len(Same) + TNFP = len(Cross) + Recall_Pos, Recall_Neg = [], [] + Precision_Pos, Precision_Neg = [], [] + Correct = [] + for th in Thresh: + TP = np.sum(Same > th) + FN = TPFN - TP + TN = np.sum(Cross < th) + FP = TNFP - TN + + Recall_Pos.append(TP/TPFN) + Recall_Neg.append(TN/TNFP) + Precision_Pos.append(TP/(TP+FP)) + Precision_Neg.append(TN/(TN+FN)) + Correct.append((TN+TP)/(TPFN+TNFP)) + + fig, ax = plt.subplots() + ax.plot(Thresh, Correct, 'r', label='Correct: (TN+TP)/(TPFN+TNFP)') + ax.plot(Thresh, Recall_Pos, 'b', label='Recall_Pos: TP/TPFN') + ax.plot(Thresh, Recall_Neg, 'g', label='Recall_Neg: TN/TNFP') + ax.plot(Thresh, Precision_Pos, 'c', label='Precision_Pos: TP/(TP+FP)') + ax.plot(Thresh, Precision_Neg, 'm', label='Precision_Neg: TN/(TN+FN)') + + ax.set_xlim([0, 1]) + ax.set_ylim([0, 1]) + ax.grid(True) + ax.set_title('PrecisePos & PreciseNeg') + ax.legend() + plt.show() + + print("Haved done!!!") + + + + + +if __name__ == '__main__': + + # Network Setup + if conf.testbackbone == 'resnet18': + # model = ResIRSE(img_size, embedding_size, conf.drop_ratio).to(device) + model = resnet18().to(device) + # elif conf.testbackbone == 'resnet34': + # model = resnet34().to(device) + # elif conf.testbackbone == 'resnet50': + # model = resnet50().to(device) + # elif conf.testbackbone == 'mobilevit_s': + # model = mobilevit_s().to(device) + # elif conf.testbackbone == 'mobilenetv3': + # model = MobileNetV3_Small().to(device) + # elif conf.testbackbone == 'mobilenet_v1': + # model = mobilenet_v1().to(device) + # elif conf.testbackbone == 'PPLCNET_x1_0': + # model = PPLCNET_x1_0().to(device) + # elif conf.testbackbone == 'PPLCNET_x0_5': + # model = PPLCNET_x0_5().to(device) + # elif conf.backbone == 'PPLCNET_x2_5': + # model = PPLCNET_x2_5().to(device) + # elif conf.testbackbone == 'mobilenet_v2': + # model = mobilenet_v2().to(device) + # elif conf.testbackbone == 'resnet14': + # model = resnet14().to(device) + else: + raise ValueError('Have not model {}'.format(conf.backbone)) + + print('load model {} '.format(conf.testbackbone)) + # model = nn.DataParallel(model).to(conf.device) + model.load_state_dict(torch.load(conf.test_model, map_location=conf.device)) + model.eval() + if not conf.group_test: + images = unique_image(conf.test_list) + images = [osp.join(conf.test_val, img) for img in images] + + groups = group_image(images, conf.test_batch_size) ##根据batch_size取图片 + + feature_dict = dict() + for group in groups: + d = featurize(group, conf.test_transform, model, conf.device) + feature_dict.update(d) + # print('feature_dict', feature_dict) + accuracy, threshold = compute_accuracy(feature_dict, conf.test_list, conf.test_val) + + print( + f"Test Model: {conf.test_model}\n" + f"Accuracy: {accuracy:.3f}\n" + f"Threshold: {threshold:.3f}\n" + ) + elif conf.group_test: + """ + conf.test_val: 测试数据集地址 + conf.test_group_json：测试数据分组配置文件 + """ + filename = conf.test_group_json + + filename = "../cl/images_1.json" + with open(filename, 'r', encoding='utf-8') as file: + content_list_read = json.load(file) + + + compute_contrast_accuracy(content_list_read) + + + + + +# ============================================================================= +# Similarity, Label = compute_group_accuracy(content_list_read) +# print('allSimilarity >> {}'.format(Similarity)) +# print('allLabel >> {}'.format(Label)) +# compute_accuracy_recall(np.array(Similarity), np.array(Label)) +# # compute_group_accuracy(data_loaded) +# +# ============================================================================= diff --git a/tracking/说明文档.txt b/tracking/说明文档.txt index 850ee9a..4b1700d 100644 --- a/tracking/说明文档.txt +++ b/tracking/说明文档.txt @@ -22,6 +22,7 @@ module_analysis.py enentmatch.py 1:n 模拟测试，have Deprecated! + contrast_analysis.py 1:n 现场测试评估。 main():