更改

.idea/CopilotChatHistory.xml

@@ -3,6 +3,12 @@
   <component name="CopilotChatHistory">
     <option name="conversations">
       <list>
+        <Conversation>
+          <option name="createTime" value="1750474299387" />
+          <option name="id" value="0197906617fb7194a0407baae2b1e2eb" />
+          <option name="title" value="新对话 2025年6月21日 10:51:39" />
+          <option name="updateTime" value="1750474299387" />
+        </Conversation>
         <Conversation>
           <option name="createTime" value="1749793513436" />
           <option name="id" value="019767d21fdc756ba782b33c8b14cdf1" />

configs/scatter.yml

@@ -18,20 +18,20 @@ models:
 
 # 训练参数
 training:
-  epochs: 300                         # 总训练轮次
+  epochs: 600                         # 总训练轮次
   batch_size: 64                     # 批次大小
-  lr: 0.005                           # 初始学习率
+  lr: 0.0004                           # 初始学习率
   optimizer: "sgd"                    # 优化器类型
   metric: 'arcface'                   # 损失函数类型（可选：arcface/cosface/sphereface/softmax）
   loss: "cross_entropy"               # 损失函数类型（可选：cross_entropy/cross_entropy_smooth/center_loss/center_loss_smooth/arcface/cosface/sphereface/softmax）
   lr_step: 10                         # 学习率调整间隔（epoch）
-  lr_decay: 0.98                      # 学习率衰减率
+  lr_decay: 0.95                      # 学习率衰减率
   weight_decay: 0.0005                  # 权重衰减
-  scheduler: "cosine_annealing"       # 学习率调度器（可选：cosine_annealing/step/none）
+  scheduler: "step"       # 学习率调度器（可选：cosine_annealing/step/none）
   num_workers: 32                     # 数据加载线程数
-  checkpoints: "./checkpoints/resnet18_scatter_6.2/" # 模型保存目录
+  checkpoints: "./checkpoints/resnet18_scatter_6.26/" # 模型保存目录
   restore: True
-  restore_model: "checkpoints/resnet18_scatter_6.2/best.pth" # 模型恢复路径
+  restore_model: "checkpoints/resnet18_scatter_6.25/best.pth" # 模型恢复路径
 
 
 
@@ -46,8 +46,8 @@ data:
   train_batch_size: 128               # 训练批次大小
   val_batch_size: 100                 # 验证批次大小
   num_workers: 32                      # 数据加载线程数
-  data_train_dir: "../data_center/scatter/train"   # 训练数据集根目录
+  data_train_dir: "../data_center/scatter/v2/train"   # 训练数据集根目录
-  data_val_dir: "../data_center/scatter/val"   # 验证数据集根目录
+  data_val_dir: "../data_center/scatter/v2/val"   # 验证数据集根目录
 
 transform:
   img_size: 224                     # 图像尺寸
@@ -59,7 +59,7 @@ transform:
 
 # 日志与监控
 logging:
-  logging_dir: "./log/2025.6.2-scatter.txt"                  # 日志保存目录
+  logging_dir: "./log/2025.6.25-scatter.txt"                  # 日志保存目录
   tensorboard: true                   # 是否启用TensorBoard
   checkpoint_interval: 30             # 检查点保存间隔（epoch）
 

configs/test.yml

@@ -13,17 +13,21 @@ base:
 # 模型配置
 models:
   backbone: 'resnet18'
-  channel_ratio: 0.75
+  channel_ratio: 1.0
-  model_path: "./checkpoints/resnet18_1009/best.pth"
+  model_path: "checkpoints/resnet18_scatter_6.26/best.pth"
   half: false                        # 是否启用半精度测试（fp16）
+  contrast_learning: false
 
 # 数据配置
 data:
   test_batch_size: 128               # 训练批次大小
   num_workers: 32                      # 数据加载线程数
-  test_dir: "../data_center/contrast_learning/contrast_test_data"   # 验证数据集根目录
+  test_dir: "../data_center/scatter/v2/val_extar"   # 验证数据集根目录
   test_group_json: "../data_center/contrast_learning/model_test_data/test/inner_group_pairs.json"
-  test_list: "../data_center/contrast_learning/contrast_test_data/cross_same.txt"
+  test_list: "../data_center/scatter/val_extar_cross_same.txt"
+  group_test: false
+  save_image_joint: true
+  image_joint_pth: "./joint_images"
 
 transform:
   img_size: 224                     # 图像尺寸

configs/transform.yml

@@ -18,8 +18,8 @@ models:
   channel_ratio: 0.75
   model_path: "../checkpoints/resnet18_1009/best.pth"
   onnx_model: "../checkpoints/resnet18_1009/best.onnx"
-  rknn_model: "../checkpoints/resnet18_1009/best_rknn2.3.2.rknn"
+  rknn_model: "../checkpoints/resnet18_1009/best_rknn2.3.2_batch16.rknn"
-  rknn_batch_size: 1
+  rknn_batch_size: 16
 
 # 日志与监控
 logging:

test_ori.py

@@ -11,6 +11,7 @@ import matplotlib.pyplot as plt
 
 # from config import config as conf
 from tools.dataset import get_transform
+from tools.image_joint import merge_imgs
 from configs import trainer_tools
 import yaml
 
@@ -22,6 +23,7 @@ embedding_size = conf["base"]["embedding_size"]
 img_size = conf["transform"]["img_size"]
 device = conf["base"]["device"]
 
+
 def unique_image(pair_list: str) -> Set[str]:
     unique_images = set()
     try:
@@ -38,7 +40,7 @@ def unique_image(pair_list: str) -> Set[str]:
     except IOError as e:
         print(f"Error reading pair list file: {e}")
         raise
-    
+
     return unique_images
 
 
@@ -56,11 +58,11 @@ def group_image(images: Set[str], batch_size: int) -> List[List[str]]:
     image_list = list(images)
     num_images = len(image_list)
     batches = []
-    
+
     for i in range(0, num_images, batch_size):
         batch_end = min(i + batch_size, num_images)
         batches.append(image_list[i:batch_end])
-        
+
     return batches
 
 
@@ -89,21 +91,21 @@ def test_preprocess(images: list, transform) -> torch.Tensor:
 
 
 def featurize(
-    images: List[str], 
+        images: List[str],
-    transform: callable,
+        transform: callable,
-    net: nn.Module,
+        net: nn.Module,
-    device: torch.device,
+        device: torch.device,
-    train: bool = False
+        train: bool = False
 ) -> Dict[str, torch.Tensor]:
     try:
         # Select appropriate preprocessing
         preprocess_fn = _preprocess if train else test_preprocess
-        
+
         # Preprocess and move to device
         data = preprocess_fn(images, transform)
         data = data.to(device)
         net = net.to(device)
-        
+
         # Extract features with automatic mixed precision
         with torch.no_grad():
             if conf['models']['half']:
@@ -111,12 +113,16 @@ def featurize(
             features = net(data)
         # Create path-to-feature mapping
         return {img: feature for img, feature in zip(images, features)}
-        
+
     except Exception as e:
         print(f"Error in feature extraction: {e}")
         raise
+
+
 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)
@@ -179,22 +185,23 @@ def compute_accuracy_recall(score, labels):
         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))
+        # 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))
         Correct.append(0 if TP == 0 else (TP + TN) / (TP + FP + TN + FN))
-
+        print("Threshold:{} PrecisePos:{},recall:{},PreciseNeg:{},recall_TN:{}".format(th, PrecisePos[-1], recall[-1],
+                                                                                       PreciseNeg[-1], recall_TN[-1]))
     showHist(Same, Cross)
     showgrid(recall, recall_TN, PrecisePos, PreciseNeg, Correct)
 
 
 def compute_accuracy(
-    feature_dict: Dict[str, torch.Tensor],
+        feature_dict: Dict[str, torch.Tensor],
-    pair_list: str,
+        pair_list: str,
-    test_root: str
+        test_root: str
 ) -> Tuple[float, float]:
     try:
         with open(pair_list, 'r') as f:
@@ -205,37 +212,43 @@ def compute_accuracy(
 
     similarities = []
     labels = []
-    
+
     for pair in pairs:
         pair = pair.strip()
         if not pair:
             continue
-            
+
-        try:
+        # try:
-            img1, img2, label = pair.split()
+        img1, img2, label = pair.split()
-            img1_path = osp.join(test_root, img1)
+        img1_path = osp.join(test_root, img1)
-            img2_path = osp.join(test_root, img2)
+        img2_path = osp.join(test_root, img2)
-            
+
-            # Verify features exist
+        # Verify features exist
-            if img1_path not in feature_dict or img2_path not in feature_dict:
+        if img1_path not in feature_dict or img2_path not in feature_dict:
-                raise ValueError(f"Missing features for image pair: {img1_path}, {img2_path}")
+            raise ValueError(f"Missing features for image pair: {img1_path}, {img2_path}")
-                
+
-            # Get features and compute similarity
+        # Get features and compute similarity
-            feat1 = feature_dict[img1_path].cpu().numpy()
+        feat1 = feature_dict[img1_path].cpu().numpy()
-            feat2 = feature_dict[img2_path].cpu().numpy()
+        feat2 = feature_dict[img2_path].cpu().numpy()
-            similarity = cosin_metric(feat1, feat2)
+        similarity = cosin_metric(feat1, feat2)
-            
+        print('{} vs {}: {}'.format(img1_path, img2_path, similarity))
-            similarities.append(similarity)
+        if conf['data']['save_image_joint']:
-            labels.append(int(label))
+            merge_imgs(img1_path,
-            
+                       img2_path,
-        except Exception as e:
+                       conf['data']['image_joint_pth'],
-            print(f"Skipping invalid pair: {pair}. Error: {e}")
+                       similarity,
-            continue
+                       label)
+        similarities.append(similarity)
+        labels.append(int(label))
+
+        # except Exception as e:
+        #     print(f"Skipping invalid pair: {pair}. Error: {e}")
+        #     continue
 
     # Find optimal threshold and accuracy
     accuracy, threshold = threshold_search(similarities, labels)
     compute_accuracy_recall(np.array(similarities), np.array(labels))
-    
+
     return accuracy, threshold
 
 
@@ -267,10 +280,10 @@ def compute_group_accuracy(content_list_read):
                 d = featurize(group[0], conf.test_transform, model, conf.device)
                 one_group_list.append(d.values())
             if data_loaded[-1] == '1':
-                similarity = deal_group_pair(one_group_list[0], one_group_list[1])
+                similarity = abs(deal_group_pair(one_group_list[0], one_group_list[1]))
                 Same.append(similarity)
             else:
-                similarity = deal_group_pair(one_group_list[0], one_group_list[1])
+                similarity = abs(deal_group_pair(one_group_list[0], one_group_list[1]))
                 Cross.append(similarity)
             allLabel.append(data_loaded[-1])
             allSimilarity.extend(similarity)
@@ -291,7 +304,17 @@ def init_model():
     print('load model {} '.format(conf['models']['backbone']))
     if torch.cuda.device_count() > 1 and conf['base']['distributed']:
         model = nn.DataParallel(model).to(conf['base']['device'])
+        ###############正常模型加载################
         model.load_state_dict(torch.load(conf['models']['model_path'], map_location=conf['base']['device']))
+        #######################################
+        ####### 对比学习模型临时运用###
+        # state_dict = torch.load(conf['models']['model_path'], map_location=conf['base']['device'])
+        # new_state_dict = {}
+        # for k, v in state_dict.items():
+        #     new_key = k.replace("module.base_model.", "module.")
+        #     new_state_dict[new_key] = v
+        # model.load_state_dict(new_state_dict, strict=False)
+        ###########################
         if conf['models']['half']:
             model.half()
         first_param_dtype = next(model.parameters()).dtype

tools/getpairs.py

@@ -188,7 +188,7 @@ class PairGenerator:
 
 
 if __name__ == "__main__":
-    original_path = '/home/lc/data_center/contrast_learning/contrast_test_data/test'
+    original_path = '/home/lc/data_center/scatter/val_extar'
     parent_dir = str(Path(original_path).parent)
     generator = PairGenerator()
 

tools/image_joint.py

@@ -0,0 +1,33 @@
+from PIL import Image, ImageDraw, ImageFont
+import os
+
+
+def merge_imgs(img1_path, img2_path, save_path, similar=None, label=None):
+        position = (50, 50)  # 文字的左上角坐标
+        color = (255, 0, 0)  # 红色文字，格式为 RGB
+        if not os.path.exists(os.sep.join([save_path, str(label)])):
+            os.makedirs(os.sep.join([save_path, str(label)]))
+        save_path = os.sep.join([save_path, str(label)])
+        img_name = os.path.basename(img1_path).split('.')[0]+'_'+os.path.basename(img2_path).split('.')[0]+'.png'
+        img1 = Image.open(img1_path)
+        img2 = Image.open(img2_path)
+        img1 = img1.resize((224,224))
+        img2 = img2.resize((224,224))
+        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))
+
+        if similar is not None:
+            similar = str(similar)+'_'+str(label)
+            draw = ImageDraw.Draw(new_img)
+            draw.text(position, str(similar), color, font_size=36)
+        os.makedirs(save_path, exist_ok=True)
+        img_save = os.path.join(save_path, img_name)
+        new_img.save(img_save)
+

train_compare.py

@@ -12,7 +12,7 @@ import matplotlib.pyplot as plt
 from configs import trainer_tools
 import yaml
 
-with open('configs/compare.yml', 'r') as f:
+with open('configs/scatter.yml', 'r') as f:
     conf = yaml.load(f, Loader=yaml.FullLoader)
 
 # Data Setup