merge_imgs

ultralytics/utils/img_joint.py

@@ -0,0 +1,50 @@
+from PIL import Image  
+import os 
+
+def merge_imgs(path1, path2, save_path):
+    for img_name in os.listdir(path1):
+        # img_dir1 = os.path.join(path1, img_dir)
+        # if os.path.isdir(img_dir1):
+        # img_dir2 = os.path.join(path2, img_dir)
+        # for img_name in os.listdir(img_dir1):
+        try:
+            img1_path = os.path.join(path1, img_name)
+            img2_path = os.path.join(path2, img_name)
+
+            img1 = Image.open(img1_path)
+            img2 = Image.open(img2_path)
+            print('img1_path',img1)
+            print('img2_path', img2)
+            assert img1.height == img2.height
+
+
+            new_img = Image.new('RGB', (img1.width + img2.width+10, img1.height))
+
+            # print('new_img', new_img)
+            new_img.paste(img1, (0, 0))
+            new_img.paste(img2, (img1.width+10, 0))
+            # save_dir = os.path.join(save_path, img_name)
+            os.makedirs(save_path, exist_ok=True)
+            img_save = os.path.join(save_path, img_name)
+            # print('img-save', img_save)
+            new_img.save(img_save)
+        except Exception as e:
+            print(e)
+            #print(img_name)
+        
+        
+#path1 = '/home/yujia/yj/yolov5-6.1/0518_cls10_v5s_new_delTP0.5/labelFn_5/'
+# path1 = 'predict5_ori_v10s/'
+# path1 = 'predict_best0524/'
+# path2 = 'predict_0613_epoch27/'
+
+# path1 = 'predict_labels110_0524/'
+# path2 = 'predict_labels110_0613/'
+
+path1 = '/home/lc/ieemoo-ai-gift/confusion_gift_cls4_0.45/FN/FN_3'
+path2 = '/home/lc/ieemoo-ai-gift/confusion_gift_cls4_0.45/allBox/allBox_1'
+
+#save_path = '/home/yujia/yj/yolov5-6.1/0518_cls10_v5s_new_delTP0.5/labelFn_5_allBox_merge/'
+save_path = '/home/lc/ieemoo-ai-gift/confusion_gift_cls4_0.45/FN/FN_3_joint/'
+# os.makedirs(save_path, exist_ok=True)
+merge_imgs(path1, path2, save_path)

ultralytics/utils/metrics_confusion_visual.py

@@ -389,6 +389,7 @@ class ConfusionMatrix:
                 cls = detect_cur_cpu[5]
                 c = int(cls)
                 label = f'{names[c]} {conf:.2f} iou:{float(iou):.2f}'
+                print(">>>>>>>>>>>>>>>>>>> label: {} C: {}".format(label, c))
                 if fp_flag:
                     annotator.box_label(xyxy, label, color=(125, 0, 125))  ##fp iou匹配上，类别错误  紫色框
                 else:
@@ -465,7 +466,7 @@ class ConfusionMatrix:
         annotators_tp = self.create_annotator(images_tp, Annotator, names)
         annotators_all = self.create_annotator(images_all, Annotator, names)
         ### 新建不同检测类别保存的文件夹，文件夹名称为类别索引
-        date_path = 'confusion_0717_cls10_' + str(self.iou_thres)
+        date_path = 'confusion_gift_cls4_' + str(self.iou_thres)
         paths_fn = self.makdirs_file(date_path, img_name, self.nc, str_c='FN')
         paths_fp = self.makdirs_file(date_path, img_name, self.nc, str_c='FP')
         paths_fp_bg = self.makdirs_file(date_path, img_name, self.nc, str_c='FP_bg')
@@ -485,7 +486,7 @@ class ConfusionMatrix:
         cls_flag_list_tp = [False for _ in range(self.nc)]  ###混淆矩阵斜对角线上类别flag
 
         ##  混淆矩阵可视化 flag 设置
-        save_oneImg = False  ### 将所有pred_box与gt_box匹配结果画在一张图片上
+        save_oneImg = True  #pred_box与gt_box匹配结果画在一张图片上
         save_byClass = True  ### 将所有pred_box与gt_box匹配结果按box类别分类保存，其中tp、fn、fp按gt_box类别划分，fp_bg按pred_box划分
         cls_flag_list = []
         ###=========================

ultralytics/utils/plotting.py

@@ -84,6 +84,7 @@ class Colors:
 
     def __call__(self, i, bgr=False):
         """Converts hex color codes to RGB values."""
+        i=0
         c = self.palette[int(i) % self.n]
         return (c[2], c[1], c[0]) if bgr else c