modify 1:1 比对方式

contrast/seqfeat_compare.py

@@ -0,0 +1,200 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Fri Aug  9 10:36:45 2024
+分析图像对间的相似度
+
+@author: ym
+"""
+import os
+import cv2
+import numpy as np
+
+import torch
+import sys
+from scipy.spatial.distance import cdist
+
+''' 加载 LC 定义的模型形式'''
+from config import config as conf
+from model import resnet18 as resnet18
+from test_ori import inference_image
+##============ load resnet mdoel 
+model = resnet18().to(conf.device)
+# model = nn.DataParallel(model).to(conf.device)
+model.load_state_dict(torch.load(conf.test_model, map_location=conf.device))
+model.eval()
+print('load model {} '.format(conf.testbackbone))
+
+IMG_FORMAT = ['.bmp', '.jpg', '.JPG', '.jpeg', '.png']
+
+# =============================================================================
+# ''' 加载REID中定义的模型形式'''
+# sys.path.append(r"D:\DetectTracking")
+# from tracking.trackers.reid.reid_interface import ReIDInterface 
+# from tracking.trackers.reid.config import config as ReIDConfig
+# ReIDEncoder = ReIDInterface(ReIDConfig)
+# 
+# def inference_image_ReID(images):
+#     batch_patches = []
+#     patches = []
+#     for d, img1 in enumerate(images):
+# 
+# 
+#         img = img1[:, :, ::-1].copy()    # the model expects RGB inputs
+#         patch = ReIDEncoder.transform(img)
+#         
+#         # patch = patch.to(device=self.device).half()
+#         if str(ReIDEncoder.device) != "cpu":
+#             patch = patch.to(device=ReIDEncoder.device).half()
+#         else:
+#             patch = patch.to(device=ReIDEncoder.device)
+#         
+#         patches.append(patch)
+#         if (d + 1) % ReIDEncoder.batch_size == 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, ReIDEncoder.embedding_size))
+#     for patches in batch_patches:
+#         pred = ReIDEncoder.model(patches)
+#         pred[torch.isinf(pred)] = 1.0
+#         feat = pred.cpu().data.numpy()
+#         features = np.vstack((features, feat))
+# 
+#     return features
+# =============================================================================
+
+
+def silimarity_compare():
+    
+    imgpaths = r"D:\DetectTracking\contrast\images\2"
+    
+    
+    filepaths = []   
+    for root, dirs, filenames in os.walk(imgpaths):  
+        for filename in filenames:
+            file, ext = os.path.splitext(filename)
+            if ext not in IMG_FORMAT: continue
+            
+            file_path = os.path.join(root, filename)  
+            filepaths.append(file_path)
+    
+    feature = inference_image(filepaths, conf.test_transform, model, conf.device)
+    feature /= np.linalg.norm(feature, axis=1)[:, None]
+    
+    similar = 1 - np.maximum(0.0, cdist(feature, feature, metric='cosine')) 
+    
+    
+    print("Done!")
+
+
+
+def similarity_compare_sequence(root_dir):
+    '''
+    root_dir：包含 "subimgs"字段的文件夹中图像为 subimg子图
+    功能：相邻帧子图间相似度比较
+    
+    
+    '''
+
+    all_files = []
+    extensions = ['.png', '.jpg']
+    for dirpath, dirnames, filenames in os.walk(root_dir):
+        filepaths = []
+        for filename in filenames:
+            if os.path.basename(dirpath).find('subimgs') < 0: 
+                continue
+            file, ext = os.path.splitext(filename)
+            if ext in extensions:
+                imgpath = os.path.join(dirpath, filename)
+                filepaths.append(imgpath)
+        nf = len(filepaths)
+        if nf==0:
+            continue
+
+        fnma = os.path.basename(filepaths[0]).split('.')[0]
+        imga = cv2.imread(filepaths[0])
+        ha, wa = imga.shape[:2]
+        
+        for i in range(1, nf):
+            fnmb = os.path.basename(filepaths[i]).split('.')[0]
+            
+            imgb = cv2.imread(filepaths[i])
+            hb, wb = imgb.shape[:2]
+            
+            
+            feats = inference_image_ReID(((imga, imgb)))
+            
+            similar = 1 - np.maximum(0.0, cdist(feats, feats, metric='cosine')) 
+            
+            
+            h, w  = max((ha, hb)), max((wa, wb))
+            img = np.zeros(((h, 2*w, 3)), np.uint8)
+            img[0:ha, 0:wa], img[0:hb, w:(w+wb)] = imga, imgb
+
+            linewidth = max(round(((h+2*w))/2 * 0.001), 2)
+            cv2.putText(img,
+                        text=f'{similar[0,1]:.2f}',     # Text string to be drawn
+                        org=(max(w-20, 10), h-10),       # Bottom-left corner of the text string
+                        fontFace=0,                     # Font type
+                        fontScale=linewidth/3,          # Font scale factor
+                        color=(0, 0, 255),              # Text color
+                        thickness=linewidth,            # Thickness of the lines used to draw a text
+                        lineType=cv2.LINE_AA,           # Line type
+                        )                           
+            spath = os.path.join(dirpath, 's'+fnma+'-vs-'+fnmb+'.png')
+            cv2.imwrite(spath, img)
+            
+            
+            fnma = os.path.basename(filepaths[i]).split('.')[0]
+            imga = imgb.copy()
+            ha, wa = imga.shape[:2]
+    
+    
+    return
+    
+    
+def main():
+    root_dir = r"D:\contrast\dataset\result\20240723-112242_6923790709882"
+    
+    try:
+        similarity_compare_sequence(root_dir)   
+    except Exception as e:
+        print(f'Error: {e}') 
+    
+
+
+
+if __name__ == '__main__':
+    # main()
+    
+    silimarity_compare()
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+