modified for site test

data.txt

@@ -0,0 +1 @@
+-0.011240 0.008260 -0.006586 -0.001030 -0.013720 -0.002563 -0.046692 0.001246 -0.004448 -0.000575 0.011596 -0.008257 0.013804 0.013636 -0.015231 0.009997 0.002080 -0.004984 -0.008263 -0.017190 0.015428 -0.023105 -0.032034 0.014638 0.017993 -0.020305 -0.026571 -0.013695 0.019300 -0.008173 -0.014615 -0.029739 -0.005307 0.004259 0.027220 -0.003702 0.014700 -0.007800 -0.003245 -0.033670 -0.046784 -0.014097 0.022587 -0.007240 0.020345 -0.007308 -0.032579 -0.026351 0.007981 -0.004793 -0.003228 -0.001727 -0.019257 -0.004215 -0.018438 0.022703 -0.007637 -0.009849 -0.039515 -0.063543 0.038563 0.024434 -0.020794 -0.013288 0.024293 0.005290 -0.023577 0.043458 -0.008223 0.009316 0.015879 -0.016333 0.008900 0.014967 0.023980 -0.007785 0.011072 0.014957 -0.014673 -0.015173 0.001238 -0.002713 -0.010376 -0.009099 -0.019973 0.036330 -0.025961 -0.004987 -0.003738 -0.029137 0.027006 0.013193 -0.013956 -0.010477 0.006946 0.003161 0.006327 -0.010057 -0.000473 0.039186 -0.017787 0.030310 0.001994 0.021185 -0.010288 -0.026407 -0.007325 -0.035419 0.013209 0.040455 0.022204 0.011113 0.013391 -0.025687 -0.008719 0.018367 -0.044993 0.000359 -0.027293 0.017411 0.016963 -0.012727 0.004216 -0.007090 0.020172 -0.007653 -0.003869 0.000472 -0.006563 -0.010175 -0.012288 0.030884 0.021227 -0.008667 0.001995 0.002351 0.001223 0.024315 0.048389 -0.016056 0.015207 0.035997 -0.017303 0.029428 -0.018798 0.009189 -0.008502 0.036859 -0.003675 -0.003153 -0.017599 -0.020731 0.023639 0.019200 0.017236 0.015245 0.000899 0.013015 -0.026410 0.003367 0.007493 0.006190 -0.008258 0.017456 0.007086 -0.015679 -0.035943 -0.028529 -0.029751 0.000321 0.027217 0.002749 -0.016362 0.003308 0.011506 0.005780 0.008492 -0.002685 -0.006707 -0.001248 -0.005391 -0.010571 0.000716 -0.015180 -0.008275 -0.002362 -0.002915 -0.011054 -0.007975 0.016847 -0.003256 0.004353 -0.015026 -0.007171 -0.019375 -0.002358 -0.029985 -0.004786 0.008605 0.000120 0.016673 -0.029609 -0.030924 0.004636 -0.022859 -0.002508 0.028345 -0.007889 0.017705 0.012368 -0.020287 -0.001889 0.008966 0.017198 0.031740 -0.016312 -0.029071 0.014328 -0.029138 -0.006111 -0.038278 -0.006854 -0.006448 -0.016257 -0.003441 -0.003229 -0.012162 -0.014835 0.011474 -0.010222 -0.017947 0.021293 -0.006472 0.003448 0.005727 -0.033055 0.005207 -0.008356 -0.015410 0.029921 -0.030446 0.000284 -0.029262 0.034003 0.017720 -0.029013 0.010341 -0.037284 0.006937 0.060924 0.015401 0.023268 0.009021 -0.028368 -0.019817 0.034878

featureVal.py

@@ -104,25 +104,76 @@ def inference_image(image, detections):
 
     return imgs, features
 
-def test_dog():
+def readimg():
+    imgpath = r"D:\datasets\ym\Img_ResnetData\result\0.png"
+    image = cv2.imread(imgpath)
     
+    img = cv2.resize(image, (224, 224))
+    
+    cv2.imwrite('0_224x224.jpg', img)
+    
+
+def readdata(datapath):
     
-    datapath = r"D:\datasets\ym\Img_ResnetData\dog_224x224\dog_224x224.txt"
     with open(datapath, 'r') as file:  
         lines = file.readlines()
     dlist = lines[0].split(',')
     dfloat = [float(d) for d in dlist]
-    
     afeat = np.array(dfloat).reshape(1, -1)
     
+    return afeat
+def readrawimg(datapath):
+    with open(datapath, 'r') as file:  
+        llines = file.readlines()
+    imgs = []
+    
+    row = 224
+    
+    for i in range(8):
+        lines = llines[i*224 : (i+1)*224]
+        
     
     
-    imgpath = r"D:\datasets\ym\Img_ResnetData\dog_224x224\dog_224x224.jpg"
+        img = np.empty((224, 224, 0), dtype=np.float32)
-    image = cv2.imread(imgpath)
+        imgr = np.empty((0, 224), dtype=np.float32)
+        imgg = np.empty((0, 224), dtype=np.float32) 
+        imgb = np.empty((0, 224), dtype=np.float32)
+        
+        for line in lines:
+            dlist = line.split(' ')[0:224]
+            
+            img_r = np.array([float(s.split(',')[0]) for s in dlist], dtype=np.float32).reshape(1, -1)
+            img_g = np.array([float(s.split(',')[1]) for s in dlist], dtype=np.float32).reshape(1, -1)
+            img_b = np.array([float(s.split(',')[2]) for s in dlist], dtype=np.float32).reshape(1, -1)
+            
+            # img_r = [float(s.split(',')[0]) for s in dlist if len(s.split(',')[0].encode('utf-8')) == 4]
+            # img_g = [float(s.split(',')[1]) for s in dlist if len(s.split(',')[1].encode('utf-8')) == 4]
+            # img_b = [float(s.split(',')[2]) for s in dlist if len(s.split(',')[2].encode('utf-8')) == 4]
+            
+            imgr = np.concatenate((imgr, img_r), axis=0)
+            imgg = np.concatenate((imgg, img_g), axis=0)
+            imgb = np.concatenate((imgb, img_b), axis=0)
+        
+        imgr = imgr[:, :, None]
+        imgg = imgg[:, :, None]
+        imgb = imgb[:, :, None]
+    
+        img = np.concatenate((imgb, imgg, imgr), axis=2).astype(np.uint8)
+        
+        imgs.append(img)
+
+    return imgs
+    
+
+
+
+
+def inference(image):
+    
     patches = []
     
-    img = image[:, :, ::-1].copy()    # the model expects RGB inputs
+    image = image[:, :, ::-1].copy()    # the model expects RGB inputs
-    patch = ReIDEncoder.transform(img)
+    patch = ReIDEncoder.transform(image)
     
     patch = patch.to(device=ReIDEncoder.device)
     
@@ -132,29 +183,43 @@ def test_dog():
     pred[torch.isinf(pred)] = 1.0
     bfeat = pred.cpu().data.numpy()
     
+    return bfeat
+    
+
+
+def test_img_feat():
+    # datapath = r"D:\datasets\ym\Img_ResnetData\aa\aa.txt"
+    # afeat = readdata(datapath)
+
+    imgpath = r"D:\datasets\ym\Img_ResnetData\aa\aa.jpg"
+    img = cv2.imread(imgpath)
+    bfeat = inference(img)
     
     
+    datapath = r"D:\datasets\ym\Img_ResnetData\rawimg\7.txt"
+    afeat = readdata(datapath)
+
+    rawpath = r"D:\datasets\ym\Img_ResnetData\rawimg\28950640607_mat_rgb"
+    imgx = readrawimg(rawpath)
+    cv2.imwrite("rawimg.png", imgx[7])
+    bfeatx = inference(imgx[7])
+    
+    cost_matrix = 1 - np.maximum(0.0, cdist(afeat, bfeatx, 'cosine'))
+    
+    imgpath1 = r"D:\datasets\ym\Img_ResnetData\result\0_224x224.png"
+    img1 = cv2.imread(imgpath1)
+    bfeat1 = inference(img1)
     
     aafeat = afeat / np.linalg.norm(afeat, ord=2, axis=1, keepdims=True)
     bbfeat = bfeat / np.linalg.norm(bfeat, ord=2, axis=1, keepdims=True)
     
-    
     cost_matrix = 1 - np.maximum(0.0, cdist(aafeat, bbfeat, 'cosine'))
-    
-    
-    
-    
-    print("Done!!!")
-    
-    
-    
-    
-    
 
+    print("Done!!!")
 
 def main():
     imgpath = r"D:\datasets\ym\Img_ResnetData\20240531-103547_0354b1cb-53fa-48de-86cd-ac3c5b127ada_6921168593576\3568800050000_0.jpeg"
-    datapath = r"D:\datasets\ym\Img_ResnetData\0_tracker_inout.data"
+    datapath = r"D:\datasets\ym\Img_ResnetData\20240531-103547_0354b1cb-53fa-48de-86cd-ac3c5b127ada_6921168593576\0_tracker_inout.data"
     savepath = r"D:\datasets\ym\Img_ResnetData\result"
     
     image = cv2.imread(imgpath)
@@ -184,9 +249,11 @@ def main():
     
     
 if __name__ == '__main__':    
-    main()
+    # main()
     
-    # test_dog()
+    # readimg()
+    
+    test_img_feat()
 
     
     

track_reid.py

@@ -266,11 +266,7 @@ def run(
                 # Rescale boxes from img_size to im0 size
                 det[:, :4] = scale_boxes(im.shape[2:], det[:, :4], im0.shape).round()
                 
-                det = det.cpu().numpy()
+                # det = det.cpu().numpy()                
-                det = np.concatenate([det[:, :4], np.arange(nd).reshape(-1, 1), det[:, 4:]], axis=-1) 
-
-                DetBoxes = np.concatenate([DetBoxes, det[:, :6]], axis=0)
-                
                 ## ============================================================ 前后帧相同 boxes 的特征赋值 
                 # def static_estimate(box1, box2, TH1=8, TH2=12):
                 #     dij_abs = max(np.abs(box1 - box2))
@@ -321,14 +317,9 @@ def run(
                 
                 '''================== 1. 存储 dets/subimgs/features Dict ============='''
                 imgs, features = inference_image(im0, tracks)
-                
 
                 TrackerFeats = np.concatenate([TrackerFeats, features], axis=0)
-                
+
-                
-                
-                
-                
                 imgdict = {}
                 boxdict = {}
                 featdict = {}
@@ -339,8 +330,7 @@ def run(
                 TracksDict[f"frame_{int(dataset.frame)}"] = {"imgs":imgdict, "boxes":boxdict, "feats":featdict}
                 
                 track_boxes = np.concatenate([track_boxes, tracks], axis=0)
-                    
+
-            
                 '''================== 2. 提取手势位置 ==================='''
                 # idx_0 = tracks[:, 6].astype(np.int_) == 0
                 # hn = 0
@@ -422,8 +412,10 @@ def run(
     trackdicts_dir = trackdicts_dir.joinpath(f'{filename}.pkl')
     with open(trackdicts_dir, 'wb') as file:
         pickle.dump(TracksDict, file)
+    
+    # np.save(f'{filename}.npy', DetBoxes)
 
-        
+    
     '''======================== 3. save hand_local data =================='''
     # handlocal_dir = Path('./tracking/data/handlocal/')
     # if not handlocal_dir.exists():
@@ -539,17 +531,16 @@ def main_loop(opt):
         #           r"D:\datasets\ym\广告板遮挡测试\8\2500441577966_20240508-175946_front_addGood_70f75407b7ae_155_17788571404.mp4"
         #           ]
         
-        # files = [r"D:\datasets\ym\视频\20240529\110518062-090ac04c-0a8c-479f-bc18-cb3553c90683-0_seek0.017962635633665514.mp4"]
+        files = [r"D:\datasets\ym\广告板遮挡测试\8\6907149227609_20240508-174733_back_returnGood_70f754088050_425_17327712807.mp4"]
-        files = [r"D:\datasets\ym\视频\20240529\110518060-550b7c4d-9946-4aa4-9131-81008692cd65-1_seek0.7670042724609232.mp4"]
         
         
         for file in files:
             optdict["source"] = file
             run(**optdict)
             
-            # k += 1
+            k += 1
-            # if k == 3:
+            if k == 1:
-            #     break
+                break
     elif os.path.isfile(p):
         optdict["source"] = p
         run(**vars(opt))

tracking/contrast_analysis.py

@@ -0,0 +1,365 @@
+import os.path
+import shutil
+
+import numpy as np
+import matplotlib.pyplot as plt
+import cv2
+from utils.plotting import Annotator, colors
+import sys
+sys.path.append(r"D:\DetectTracking")
+from tracking.utils.read_data import extract_data, read_deletedBarcode_file, read_tracking_output
+from tracking.utils.plotting import draw_tracking_boxes
+
+def showHist(err, correct):
+    err = np.array(err)
+    correct = np.array(correct)
+
+    fig, axs = plt.subplots(2, 1)
+    axs[0].hist(err, bins=50, edgecolor='black')
+    axs[0].set_xlim([0, 1])
+    axs[0].set_title('err')
+
+    axs[1].hist(correct, bins=50, edgecolor='black')
+    axs[1].set_xlim([0, 1])
+    axs[1].set_title('correct')
+    plt.show()
+
+def showgrid(recall, prec, ths):
+    # x = np.linspace(start=-0, stop=1, num=11, endpoint=True).tolist()
+    fig = plt.figure(figsize=(10, 6))
+    plt.plot(ths, recall, color='red', label='recall')
+    plt.plot(ths, prec, color='blue', label='PrecisePos')
+    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_recall_precision(err_similarity, correct_similarity):
+    ths = np.linspace(0, 1, 11)
+    recall, prec = [], []
+    for th in ths:
+        TP = len([num for num in correct_similarity if num >= th])
+        FP = len([num for num in err_similarity if num >= th])
+        if (TP+FP) == 0:
+            prec.append(1)
+            recall.append(0)
+        else:
+            prec.append(TP / (TP + FP))
+            recall.append(TP / (len(err_similarity) + len(correct_similarity)))
+            
+    showgrid(recall, prec, ths)
+    return recall, prec
+
+
+# =============================================================================
+# def read_tracking_output(filepath):
+#     boxes = []
+#     feats = []
+#     with open(filepath, 'r', encoding='utf-8') as file:
+#         for line in file:  
+#             line = line.strip()  # 去除行尾的换行符和可能的空白字符
+#             
+#             if not line:  
+#                 continue 
+#             
+#             if line.endswith(','):  
+#                 line = line[:-1]
+#                 
+#             data = np.array([float(x) for x in line.split(",")])
+#             if data.size == 9:
+#                 boxes.append(data)
+#             if data.size == 256:
+#                 feats.append(data)
+#     
+#     return np.array(boxes), np.array(feats) 
+# =============================================================================
+
+def read_tracking_imgs(imgspath):
+    '''
+    input:
+        imgspath：该路径中的图像为Yolo算法的输入图像，640x512
+    output：
+        imgs_0：后摄图像，根据 frameId 进行了排序
+        imgs_1：前摄图像，根据 frameId 进行了排序 
+    '''
+    imgs_0, frmIDs_0, imgs_1, frmIDs_1 = [], [], [], []
+    
+    for filename in os.listdir(imgspath):
+        file, ext = os.path.splitext(filename)
+        flist = file.split('_')
+        if len(flist)==4 and ext==".jpg":
+            camID, frmID = flist[0], int(flist[-1])
+            imgpath = os.path.join(imgspath, filename)
+            img = cv2.imread(imgpath)
+            
+            if camID=='0':
+                imgs_0.append(img)
+                frmIDs_0.append(frmID)
+            if camID=='1':
+                imgs_1.append(img)
+                frmIDs_1.append(frmID)
+                
+    if len(frmIDs_0):
+        indice = np.argsort(np.array(frmIDs_0))
+        imgs_0 = [imgs_0[i] for i in indice ]
+    if len(frmIDs_1):
+        indice = np.argsort(np.array(frmIDs_1))
+        imgs_1 = [imgs_1[i] for i in indice ]
+    
+    return imgs_0, imgs_1
+    
+    
+# =============================================================================
+# def draw_tracking_boxes(imgs, tracks):
+#     '''tracks: [x1, y1, x2, y2, track_id, score, cls, frame_index, box_index]
+#                 0   1   2   3      4        5     6       7           8
+#        关键：imgs中的次序和 track 中的 fid 对应
+#     ''' 
+#     subimgs = []
+#     for *xyxy, tid, conf, cls, fid, bid in tracks:                  
+#         label = f'id:{int(tid)}_{int(cls)}_{conf:.2f}'
+#         
+#         annotator = Annotator(imgs[int(fid-1)].copy())  
+#         if cls==0:
+#             color = colors(int(cls), True)
+#         elif tid>0 and cls!=0:
+#             color = colors(int(tid), True)
+#         else:
+#             color = colors(19, True)     # 19为调色板的最后一个元素
+#         
+#         pt2 = [p/2 for p in xyxy]
+#         annotator.box_label(pt2, label, color=color)
+#         img0 = annotator.result()
+#         
+#         subimgs.append(img0)
+#         
+#     return subimgs
+# =============================================================================
+
+def get_contrast_paths(pair, basepath):
+    assert(len(pair)==2 or len(pair)==3), "pair: seqdir, delete, barcodes"
+    
+    getout_fold = pair[0]       # 取出操作对应的文件夹
+    relvt_barcode = pair[1]     # 取出操作对应放入操作的 Barcode
+    if len(pair)==3:
+        error_match = pair[2]       # 取出操作错误匹配的 Barcode
+    else:
+        error_match = ''
+        
+    
+    getoutpath, inputpath, errorpath = '', '', ''
+
+    day, hms = getout_fold.strip('_').split('-')
+    
+    input_folds, times = [], []
+    errmatch_folds, errmatch_times = [], []
+    for pathname in os.listdir(basepath):
+        if pathname.endswith('_'): continue
+        if os.path.isfile(os.path.join(basepath, pathname)):continue
+        infold = pathname.split('_')
+        if len(infold)!=2: continue  
+    
+        day1, hms1 = infold[0].split('-') 
+        
+        if day1==day and infold[1]==relvt_barcode and int(hms1)<int(hms):
+            input_folds.append(pathname)
+            times.append(int(hms1))
+        
+        if day1==day and len(error_match) and infold[1]==error_match and int(hms1)<int(hms):
+            errmatch_folds.append(pathname)
+            errmatch_times.append(int(hms1))
+
+    ''' 根据时间排序，选择离取出操作最近时间的文件夹，
+        作为取出操作应正确匹配的放入操作所对应的文件夹 '''
+    if len(input_folds):
+        indice = np.argsort(np.array(times)) 
+        input_fold = input_folds[indice[-1]]
+        inputpath = os.path.join(basepath, input_fold)
+    
+    '''取出操作错误匹配的放入操作对应的文件夹'''
+    if len(errmatch_folds):
+        indice = np.argsort(np.array(errmatch_times)) 
+        errmatch_fold = errmatch_folds[indice[-1]]
+        errorpath = os.path.join(basepath, errmatch_fold)
+    
+    '''放入事件文件夹地址、取出事件文件夹地址'''
+    getoutpath = os.path.join(basepath, getout_fold)
+    
+    
+    return getoutpath, inputpath, errorpath
+
+
+def save_tracking_imgpairs(pair, basepath, savepath):
+    '''
+    basepath: 原始测试数据文件夹的路径
+    savepath: 保存的目标文件夹
+    '''
+    
+    getoutpath, inputpath, errorpath = get_contrast_paths(pair, basepath)
+    
+    if len(inputpath)==0:
+        return
+
+
+    '''==== 读取放入、取出事件对应的 Yolo输入的前后摄图像，0：后摄，1：前摄 ===='''
+        
+    
+    '''==== 读取放入、取出事件对应的 tracking 输出：boxes, feats ===='''
+    
+    if len(inputpath):
+        imgs_input_0, imgs_input_1 = read_tracking_imgs(inputpath)
+        
+        input_data_0 = os.path.join(inputpath, '0_tracking_output.data')
+        input_data_1 = os.path.join(inputpath, '1_tracking_output.data')
+        boxes_input_0, feats_input_0 = read_tracking_output(input_data_0)
+        boxes_input_1, feats_input_1 = read_tracking_output(input_data_1)
+        ImgsInput_0 = draw_tracking_boxes(imgs_input_0, boxes_input_0)
+        ImgsInput_1 = draw_tracking_boxes(imgs_input_1, boxes_input_1)
+    
+    if len(getoutpath):
+        imgs_getout_0, imgs_getout_1 = read_tracking_imgs(getoutpath)
+        
+        getout_data_0 = os.path.join(getoutpath, '0_tracking_output.data')
+        getout_data_1 = os.path.join(getoutpath, '1_tracking_output.data')
+        boxes_output_0, feats_output_0 = read_tracking_output(getout_data_0)
+        boxes_output_1, feats_output_1 = read_tracking_output(getout_data_1)
+        ImgsGetout_0 = draw_tracking_boxes(imgs_getout_0, boxes_output_0)
+        ImgsGetout_1 = draw_tracking_boxes(imgs_getout_1, boxes_output_1)
+    
+    if len(errorpath):
+        
+        imgs_error_0, imgs_error_1 = read_tracking_imgs(errorpath)
+        
+        error_data_0 = os.path.join(errorpath, '0_tracking_output.data')
+        error_data_1 = os.path.join(errorpath, '1_tracking_output.data')
+        boxes_error_0, feats_error_0 = read_tracking_output(error_data_0)
+        boxes_error_1, feats_error_1 = read_tracking_output(error_data_1)
+        ImgsError_0 = draw_tracking_boxes(imgs_error_0, boxes_error_0)
+        ImgsError_1 = draw_tracking_boxes(imgs_error_1, boxes_error_1)
+        
+
+    savedir = pair[0] + pair[1]
+    if len(errorpath):
+        savedir = savedir + '_' + errorpath.split('_')[-1]
+    foldname = os.path.join(savepath, 'imgpairs', savedir)    
+    if not os.path.exists(foldname):
+        os.makedirs(foldname)
+
+    for i, img in enumerate(ImgsInput_0):
+        imgpath = os.path.join(foldname, f'input_0_{i}.png')
+        cv2.imwrite(imgpath, img)
+    for i, img in enumerate(ImgsInput_1):
+        imgpath = os.path.join(foldname, f'input_1_{i}.png')
+        cv2.imwrite(imgpath, img)
+    for i, img in enumerate(ImgsGetout_0):
+        imgpath = os.path.join(foldname, f'getout_0_{i}.png')
+        cv2.imwrite(imgpath, img)
+    for i, img in enumerate(ImgsGetout_1):
+        imgpath = os.path.join(foldname, f'getout_1_{i}.png')
+        cv2.imwrite(imgpath, img)
+        
+    for i, img in enumerate(ImgsError_0):
+        imgpath = os.path.join(foldname, f'errMatch_0_{i}.png')
+        cv2.imwrite(imgpath, img)
+    for i, img in enumerate(ImgsError_1):
+        imgpath = os.path.join(foldname, f'errMatch_1_{i}.png')
+        cv2.imwrite(imgpath, img)
+
+
+def performance_evaluate(all_list, isshow=False):
+    
+    corrpairs, correct_barcode_list, correct_similarity, errpairs, err_barcode_list,  err_similarity = [], [], [], [], [], []
+    for s_list in all_list:
+        seqdir = s_list['SeqDir'].strip()
+        delete = s_list['Deleted'].strip()
+        barcodes = [s.strip() for s in s_list['barcode']]
+        similarity = [float(s.strip()) for s in s_list['similarity']]
+        
+        if delete in barcodes[:1]:
+            corrpairs.append((seqdir, delete))
+            correct_barcode_list.append(delete)
+            correct_similarity.append(similarity[0])
+        else:
+            errpairs.append((seqdir, delete, barcodes[0]))
+            err_barcode_list.append(delete)
+            err_similarity.append(similarity[0])
+
+    
+    '''3. 计算比对性能 '''
+    if isshow:
+        compute_recall_precision(err_similarity, correct_similarity)
+        showHist(err_similarity, correct_similarity)
+        
+    return errpairs, corrpairs, err_similarity, correct_similarity
+
+
+
+def contrast_analysis(del_barcode_file, basepath, savepath, saveimgs=False):
+    '''
+    del_barcode_file: 测试数据文件，利用该文件进行算法性能分析
+    
+    '''
+
+    '''1. 读取 deletedBarcode 文件 '''
+    all_list = read_deletedBarcode_file(del_barcode_file)
+    
+    
+    '''2. 算法性能评估，并输出 (取出，删除, 错误匹配) 对 '''
+    errpairs, corrpairs, _, _ = performance_evaluate(all_list)
+
+    '''3. 获取 (取出，删除, 错误匹配) 对应路径，保存相应轨迹图像'''
+    relative_paths = []
+    for errpair in errpairs:
+        GetoutPath, InputPath, ErrorPath = get_contrast_paths(errpair, basepath)
+        relative_paths.append((GetoutPath, InputPath, ErrorPath))
+    
+        if saveimgs:
+            save_tracking_imgpairs(errpair, basepath, savepath)
+    
+    
+
+    
+    return relative_paths
+    
+def main():
+    del_barcode_file = 'D:/contrast/dataset/compairsonResult/deletedBarcode_20240709_pm.txt'
+    basepath = r'D:\contrast\dataset\1_to_n\709'
+    savepath = r'D:\contrast\dataset\result'
+    
+    try:
+        relative_path = contrast_analysis(del_barcode_file, basepath, savepath)
+    except Exception as e:
+        print(f'Error Type: {e}')
+    
+
+
+    
+
+
+if __name__ == '__main__':
+    main()
+    
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

tracking/data/说明文档.txt

@@ -0,0 +1 @@
+文件夹 trackdicts_20240608 和 trackdicts_1 下的数据为和手部关联前的跟踪结果数据

tracking/dotrack/dotracks.py

@@ -94,6 +94,7 @@ class Track:
         self.cls = int(boxes[0, 6])
         self.frnum = boxes.shape[0]
         self.imgBorder = False
+        self.isCornpoint = False
         self.imgshape = imgshape
         self.state = MoveState.Unknown
         
@@ -101,9 +102,13 @@ class Track:
         self.start_fid = int(np.min(boxes[:, 7]))
         self.end_fid = int(np.max(boxes[:, 7]))
         
+        ''''''
         self.Hands = []
+        
         self.HandsIou = []
         
+        self.Goods = []
+        self.GoodsIou = []
         
         
         '''5个关键点（中心点、左上点、右上点、左下点、右下点 ）坐标'''
@@ -113,7 +118,7 @@ class Track:
           （中心点、左上点、右上点、左下点、右下点 ）轨迹特征'''
         self.compute_cornpts_feats()
 
-
+        '''应计算各个角点面积、平均面积'''
         mw, mh = np.mean(boxes[:, 2]-boxes[:, 0]), np.mean((boxes[:, 3]-boxes[:, 1]))
         self.mwh = np.mean((mw, mh))
         self.Area = mw * mh

tracking/dotrack/dotracks_back.py

@@ -55,6 +55,7 @@ class doBackTracks(doTracks):
         # tracks = self.sub_tracks(tracks, out_trcak)
 
         
+        [self.associate_with_hand(htrack, gtrack) for htrack in hand_tracks for gtrack in tracks]
         '''轨迹循环归并'''
         # merged_tracks = self.merge_tracks(tracks)
         merged_tracks = self.merge_tracks_loop(tracks)
@@ -66,17 +67,28 @@ class doBackTracks(doTracks):
         self.Static.extend(static_tracks)
         
         tracks = self.sub_tracks(tracks, static_tracks)
-        
-        for gtrack in tracks:
-            # print(f"Goods ID:{gtrack.tid}")
-            for htrack in hand_tracks:
-                hand_ious = self.associate_with_hand(htrack, gtrack)
-                if len(hand_ious):
-                    gtrack.Hands.append(htrack)
-                    gtrack.HandsIou.append(hand_ious)
-                    
-        self.Residual = tracks
 
+        # for gtrack in tracks:
+        #     for htrack in hand_tracks:
+        #         hand_ious = self.associate_with_hand(htrack, gtrack)
+                # if len(hand_ious):
+                #     gtrack.Hands.append(htrack)
+                #     gtrack.HandsIou.append(hand_ious)    
+                #     htrack.Goods.append((gtrack, hand_ious))
+                       
+        # for htrack in  hand_tracks:
+        #     self.merge_based_hands(htrack)
+       
+        self.Residual = tracks
+    
+    # def merge_based_hands(self, htrack):
+    #     gtracks = htrack.Goods
+        
+    #     if len(gtracks) >= 2:
+    #         atrack, afious = gtracks[0]
+    #         btrack, bfious = gtracks[1]
+    
+    
     def associate_with_hand(self, htrack, gtrack):
         ''' 
         迁移至基类：
@@ -91,6 +103,7 @@ class doBackTracks(doTracks):
         
         hboxes = np.empty(shape=(0, 9), dtype = np.float)
         gboxes = np.empty(shape=(0, 9), dtype = np.float)
+
         
         # start, end 为索引值，需要 start:(end+1)
         for start, end in htrack.moving_index:
@@ -99,18 +112,17 @@ class doBackTracks(doTracks):
             gboxes = np.concatenate((gboxes, gtrack.boxes[start:end+1, :]), axis=0)
         
         hfids, gfids = hboxes[:, 7], gboxes[:, 7]
-        fids = set(hfids).intersection(set(gfids))
+        fids = sorted(set(hfids).intersection(set(gfids)))
         
         
         if len(fids)==0:
-            return hand_ious
+            return None
-        
+
-        
         # print(f"Goods ID: {gtrack.tid}, Hand ID: {htrack.tid}")
         
         for f in fids:
-            h = np.where(hfids==f)[0][0]
+            h = np.where(hboxes[:,7] == f)[0][0]
-            g = np.where(gfids==f)[0][0]
+            g = np.where(gboxes[:,7] == f)[0][0]
             
             x11, y11, x12, y12 = hboxes[h, 0:4]
             x21, y21, x22, y22 = gboxes[g, 0:4]
@@ -124,10 +136,11 @@ class doBackTracks(doTracks):
             
             iou = union / (area1 + area2 - union + 1e-6)
             
-            if iou>0: 
+            if iou >= 0.01:                 
-                hand_ious.append((f, iou))
+                gtrack.Hands.append((htrack.tid, f, iou))
+               
         
-        return hand_ious
+        return gtrack.Hands
         
     def merge_tracks(self, Residual):
         """

tracking/dotrack/dotracks_front.py

@@ -44,21 +44,25 @@ class doFrontTracks(doTracks):
         
         '''剔除静止目标后的 tracks'''
         tracks = self.sub_tracks(tracks, static_tracks)
-         
+        
+        [self.associate_with_hand(htrack, gtrack) for htrack in hand_tracks for gtrack in tracks]
         '''轨迹循环归并'''
         merged_tracks = self.merge_tracks_loop(tracks)
         
         tracks = [t for t in merged_tracks if t.frnum > 1]
 
-        for gtrack in tracks:
+        # for gtrack in tracks:
-            # print(f"Goods ID:{gtrack.tid}")
+        #     # print(f"Goods ID:{gtrack.tid}")
-            for htrack in hand_tracks:
+        #     for htrack in hand_tracks:
-                hand_ious = self.associate_with_hand(htrack, gtrack)
+        #         hand_ious = self.associate_with_hand(htrack, gtrack)
-                if len(hand_ious):
+        #         if len(hand_ious):
-                    gtrack.Hands.append(htrack)
+        #             gtrack.Hands.append(htrack)
-                    gtrack.HandsIou.append(hand_ious)
+        #             gtrack.HandsIou.append(hand_ious)
-                    
                     
+        '''静止 tracks 判断与剔除静止 tracks'''
+        static_tracks = [t for t in tracks if t.frnum>1 and t.is_static()] 
+        tracks = self.sub_tracks(tracks, static_tracks)            
+
         freemoved_tracks = [t for t in tracks if t.is_free_move()]
         
         tracks = self.sub_tracks(tracks, freemoved_tracks)
@@ -73,10 +77,8 @@ class doFrontTracks(doTracks):
             a. 运动帧的帧索引有交集
             b. 帧索引交集部分iou均大于0
         '''
-        
         assert htrack.cls==0 and gtrack.cls!=0 and gtrack.cls!=9, 'Track cls is Error!'
         
-        hand_ious = []
         hboxes = np.empty(shape=(0, 9), dtype = np.float)
         gboxes = np.empty(shape=(0, 9), dtype = np.float)
         
@@ -87,14 +89,12 @@ class doFrontTracks(doTracks):
             gboxes = np.concatenate((gboxes, gtrack.boxes[start:end+1, :]), axis=0)
         
         hfids, gfids = hboxes[:, 7], gboxes[:, 7]
-        fids = set(hfids).intersection(set(gfids))
+        fids = sorted(set(hfids).intersection(set(gfids)))
         
         if len(fids)==0:
-            return hand_ious
+            return None
-        
+
-        
         # print(f"Goods ID: {gtrack.tid}, Hand ID: {htrack.tid}")
-        ious = []
         for f in fids:
             h = np.where(hfids==f)[0][0]
             g = np.where(gfids==f)[0][0]
@@ -111,10 +111,10 @@ class doFrontTracks(doTracks):
             
             iou = union / (area1 + area2 - union + 1e-6)
             
-            if iou>0: 
+            if iou >= 0.01: 
-                hand_ious.append((f, iou))
+                gtrack.Hands.append((htrack.tid, f, iou))
-        
+
-        return hand_ious
+        return gtrack.Hands
     
     
     

tracking/goodmatch.py

@@ -30,6 +30,26 @@ from utils.drawtracks import plot_frameID_y2, draw_all_trajectories
 
 from utils.mergetrack import readDict
 
+import csv
+
+
+def read_csv_file():
+    
+    file_path = r'D:\DeepLearning\yolov5_track\tracking\matching\featdata\Similarity.csv'   
+    with open(file_path, mode='r', newline='') as file:    
+        data = list(csv.reader(file)) 
+    
+    matrix = []
+    for i in range(1, len(data)):
+        matrix.append(data[i][1:])
+        
+    matrix = np.array(matrix, dtype = np.float32)   
+    
+    simil = 1 + (matrix-1)/2
+
+    print("done!!!")
+
+
 
 
 def get_img_filename(imgpath = r'./matching/images/' ):
@@ -747,7 +767,7 @@ def main():
     # imgsample_cleaning()
     
     '''3.1 计算事件间相似度: 将 front、back 的所有 track 特征合并'''
-    # calculate_similarity()
+    calculate_similarity()
     
     '''3.2 计算事件间相似度: 考虑前后摄的不同组合，或 track 间的不同组合'''
     # calculate_similarity_track()
@@ -766,8 +786,29 @@ def main():
     
                
 if __name__ == "__main__":
-    save_dir = Path(f'./result/') 
+    # save_dir = Path(f'./result/') 
+    # read_csv_file()
+    
     main()
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
 
 
                 

tracking/module_analysis.py

@@ -0,0 +1,390 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Thu May 30 14:03:03 2024
+
+现场测试性能分析
+
+@author: ym
+"""
+import os
+import cv2
+import numpy as np
+from pathlib import Path
+import sys
+sys.path.append(r"D:\DetectTracking")
+
+
+from tracking.utils.plotting import Annotator, colors, draw_tracking_boxes
+from tracking.utils import Boxes, IterableSimpleNamespace, yaml_load
+from tracking.trackers import BOTSORT, BYTETracker
+from tracking.dotrack.dotracks_back import doBackTracks
+from tracking.dotrack.dotracks_front import doFrontTracks 
+from tracking.utils.drawtracks import plot_frameID_y2, draw_all_trajectories
+
+from tracking.utils.read_data import extract_data, read_deletedBarcode_file, read_tracking_output
+
+from contrast_analysis import contrast_analysis
+
+from tracking.utils.annotator import TrackAnnotator
+
+W, H = 1024, 1280
+Mode = 'front'  #'back'
+ImgFormat = ['.jpg', '.jpeg', '.png', '.bmp']
+
+def video2imgs(path):
+    vpath = os.path.join(path, "videos")
+    
+    k = 0
+    have = False
+    for filename in os.listdir(vpath):
+        file, ext = os.path.splitext(filename)
+        imgdir = os.path.join(path, file)
+        if os.path.exists(imgdir): 
+            continue          
+        else:
+            os.mkdir(imgdir)
+
+        vfile = os.path.join(vpath, filename)
+        cap = cv2.VideoCapture(vfile)
+        i = 0
+        while True:
+            ret, frame = cap.read()
+            if not ret:
+                break
+            
+            i += 1
+            imgp = os.path.join(imgdir, file+f"_{i}.png")
+            cv2.imwrite(imgp, frame)
+        
+            print(filename+f": {i}")
+        
+        
+        cap.release()
+        
+        k+=1
+        if k==1000:
+            break
+
+def draw_boxes():
+    datapath = r'D:\datasets\ym\videos_test\20240530\1_tracker_inout(1).data'
+    VideosData = read_tracker_input(datapath)
+    
+    bboxes = VideosData[0][0]
+    ffeats = VideosData[0][1]
+    
+    videopath = r"D:\datasets\ym\videos_test\20240530\134458234-1cd970cf-f8b9-4e80-9c2e-7ca3eec83b81-1_seek0.10415589124891511.mp4"
+    
+    cap = cv2.VideoCapture(videopath)
+    i = 0
+    while True:
+        ret, frame = cap.read()
+        if not ret:
+            break
+        
+        
+        annotator = Annotator(frame.copy(), line_width=3)
+        
+        
+        boxes = bboxes[i]
+            
+        for *xyxy, conf, cls in reversed(boxes):
+            label = f'{int(cls)}: {conf:.2f}'                   
+   
+            color = colors(int(cls), True)
+            annotator.box_label(xyxy, label, color=color)
+            
+        img = annotator.result()    
+    
+        imgpath = r"D:\datasets\ym\videos_test\20240530\result\int8_front\{}.png".format(i+1)
+        cv2.imwrite(imgpath, img)
+    
+        print(f"Output: {i}")
+        i += 1
+    cap.release()
+    
+def read_imgs(imgspath, CamerType):
+    imgs, frmIDs = [], []
+    for filename in os.listdir(imgspath):
+        file, ext = os.path.splitext(filename)
+        flist = file.split('_')
+        if len(flist)==4 and ext in ImgFormat:
+            camID, frmID = flist[0], int(flist[-1])
+            imgpath = os.path.join(imgspath, filename)
+            img = cv2.imread(imgpath)
+            
+            if camID==CamerType:
+                imgs.append(img)
+                frmIDs.append(frmID)
+
+    if len(frmIDs):
+        indice = np.argsort(np.array(frmIDs))
+        imgs = [imgs[i] for i in indice]
+        
+    return imgs
+    
+    
+    
+    pass
+    
+    
+
+def init_tracker(tracker_yaml = None, bs=1):
+    """
+    Initialize tracker for object tracking during prediction.
+    """  
+    TRACKER_MAP = {'bytetrack': BYTETracker, 'botsort': BOTSORT}   
+    cfg = IterableSimpleNamespace(**yaml_load(tracker_yaml))
+
+    tracker = TRACKER_MAP[cfg.tracker_type](args=cfg, frame_rate=30)
+
+    return tracker
+
+def tracking(bboxes, ffeats):
+    tracker_yaml = r"./trackers/cfg/botsort.yaml"
+    tracker = init_tracker(tracker_yaml)
+    
+    TrackBoxes = np.empty((0, 9), dtype = np.float32)
+    TracksDict = {}
+    
+    '''========================== 执行跟踪处理 ============================='''
+    # dets 与 feats 应保持严格对应
+    for dets, feats in zip(bboxes, ffeats):
+        det_tracking = Boxes(dets).cpu().numpy()
+        tracks = tracker.update(det_tracking, features=feats)
+        
+        
+        '''tracks: [x1, y1, x2, y2, track_id, score, cls, frame_index, box_index]
+                    0   1   2   3      4        5     6       7           8
+                  这里，frame_index 也可以用视频的 帧ID 代替, box_index 保持不变
+        '''
+        
+        if len(tracks):
+            TrackBoxes = np.concatenate([TrackBoxes, tracks], axis=0)
+            
+            FeatDict = {}
+            for track in tracks:
+                tid = int(track[8])
+                FeatDict.update({tid: feats[tid, :]})
+                
+            frameID = tracks[0, 7]
+            
+            # print(f"frameID: {int(frameID)}")
+            assert len(tracks) == len(FeatDict), f"Please check the func: tracker.update() at frameID({int(frameID)})"
+
+            TracksDict[f"frame_{int(frameID)}"] = {"feats":FeatDict}
+            
+            
+    return TrackBoxes, TracksDict
+
+
+
+def do_tracker_tracking(fpath, save_dir):  
+    bboxes, ffeats, trackerboxes, tracker_feat_dict, trackingboxes, tracking_feat_dict = extract_data(fpath)
+    tboxes, feats_dict = tracking(bboxes, ffeats)
+    
+    CamerType = os.path.basename(fpath).split('_')[0]
+    dirname  = os.path.split(os.path.split(fpath)[0])[1]
+    if  CamerType == '1':
+        vts = doFrontTracks(tboxes, feats_dict)
+        vts.classify()
+        
+        plt = plot_frameID_y2(vts)
+        plt.savefig('front_y2.png')  
+        # plt.close()
+    elif CamerType == '0':
+        vts = doBackTracks(tboxes, feats_dict)
+        vts.classify()
+        
+        filename = dirname+'_' + CamerType
+        edgeline = cv2.imread("./shopcart/cart_tempt/edgeline.png")
+        draw_all_trajectories(vts, edgeline, save_dir, filename)
+    else:
+        print("Please check data file!")
+    
+
+
+def do_tracking(fpath, savedir):
+    '''
+    fpath: 算法各模块输出的data文件地址，匹配
+    savedir: 对 fpath 各模块输出的复现
+    分析具体视频时，需指定 fpath 和 savedir
+    '''
+    # fpath = r'D:\contrast\dataset\1_to_n\709\20240709-102758_6971558612189\1_track.data'
+    # savedir = r'D:\contrast\dataset\result\20240709-102843_6958770005357_6971558612189\error_6971558612189'
+
+    imgpath, dfname = os.path.split(fpath)
+    CamerType = dfname.split('_')[0]
+
+    bboxes, ffeats, trackerboxes, tracker_feat_dict, trackingboxes, tracking_feat_dict = extract_data(fpath)
+    
+    tracking_output_path = os.path.join(imgpath, CamerType + '_tracking_output.data')
+    tracking_output_boxes, _ = read_tracking_output(tracking_output_path)
+    
+
+    
+
+    save_dir, basename  = os.path.split(savedir)
+    if not os.path.exists(savedir):
+        os.makedirs(savedir) 
+        
+    ''' 读取 fpath 中 track.data 文件对应的图像 '''
+    
+    imgs = read_imgs(imgpath, CamerType)
+    
+    ''' 在 imgs 上画框并保存，如果 trackerboxes 的帧数和 imgs 数不匹配，返回原图'''
+    imgs_dw = draw_tracking_boxes(imgs, trackerboxes)
+    
+    if len(imgs_dw)==0:
+        imgs_dw = [img for img in imgs]
+        print(f"fpath: {imgpath}, savedir: {savedir}。Tracker输出的图像数和 imgs 中图像数不相等，无法一一匹配并画框")
+
+    for i in range(len(imgs_dw)):
+        img_savepath = os.path.join(savedir, CamerType + "_" + f"{i}.png")
+        # img = imgs_dw[i]
+        cv2.imwrite(img_savepath, imgs_dw[i])
+    
+    if not isinstance(savedir, Path): 
+        savedir = Path(savedir)
+    save_dir = savedir.parent
+    
+
+    traj_graphic = basename + '_' + CamerType
+    if  CamerType == '1':
+        vts = doFrontTracks(trackerboxes, tracker_feat_dict)
+        vts.classify()
+        
+        plt = plot_frameID_y2(vts)
+        ftpath = save_dir.joinpath(f"{traj_graphic}_front_y2.png")
+        plt.savefig(str(ftpath))  
+        plt.close()
+    elif CamerType == '0':
+        vts = doBackTracks(trackerboxes, tracker_feat_dict)
+        vts.classify()
+
+        edgeline = cv2.imread("./shopcart/cart_tempt/edgeline.png")
+        draw_all_trajectories(vts, edgeline, save_dir, traj_graphic)
+    else:
+        print("Please check data file!")
+
+
+    '''================== 现场测试的 tracking() 算法输出 =================='''
+    if CamerType == '1':
+        aline = cv2.imread("./shopcart/cart_tempt/board_ftmp_line.png")
+    elif CamerType == '0':
+        aline = cv2.imread("./shopcart/cart_tempt/edgeline.png")
+    else:
+        print("Please check data file!")
+    
+    bline = aline.copy()
+    
+    annotator = TrackAnnotator(aline, line_width=2)
+    for track in trackingboxes:
+        annotator.plotting_track(track)
+    aline = annotator.result() 
+    
+    annotator = TrackAnnotator(bline, line_width=2) 
+    if not isinstance(tracking_output_boxes, list):
+        tracking_output_boxes = [tracking_output_boxes]
+        
+    for track in tracking_output_boxes:
+        annotator.plotting_track(track)
+    bline = annotator.result()
+    
+    abimg = np.concatenate((aline, bline), axis = 1)
+    abH, abW = abimg.shape[:2]
+    cv2.line(abimg, (int(abW/2), 0), (int(abW/2), abH), (128, 255, 128), 2)
+    
+    algpath = save_dir.joinpath(f"{traj_graphic}_Alg.png")
+    cv2.imwrite(algpath, abimg)
+    
+    return    
+    
+
+def main_loop():
+    del_barcode_file = 'D:/contrast/dataset/compairsonResult/deletedBarcode_20240709_pm.txt'
+    basepath = r'D:\contrast\dataset\1_to_n\709'   # 测试数据文件夹地址
+    SavePath = r'D:\contrast\dataset\result'       # 结果保存地址
+    prefix = ["getout_", "input_", "error_"]
+    
+    
+    '''获取性能测试数据相关路径'''
+    relative_paths = contrast_analysis(del_barcode_file, basepath, SavePath)
+    
+    '''开始循环执行每次测试过任务'''
+    k = 0
+    for tuple_paths in relative_paths:
+        
+        '''生成文件夹存储结果图像的文件夹'''
+        namedirs = []
+        for data_path in tuple_paths:
+            base_name = os.path.basename(data_path).strip().split('_')
+            if len(base_name[-1]):
+                name = base_name[-1]
+            else:
+                name = base_name[0]
+            namedirs.append(name)
+
+        sdir = "_".join(namedirs)
+        savepath = os.path.join(SavePath, sdir)        
+        if not os.path.exists(savepath):
+            os.makedirs(savepath)    
+ 
+        for path in tuple_paths:
+            for filename in os.listdir(path):
+                fpath = os.path.join(path, filename)
+                
+                if os.path.isfile(fpath) and filename.find("track.data")>0:
+                    enent_name = ''
+                    
+                    '''构建结果保存文件名前缀'''
+                    for i, name in enumerate(namedirs):
+                        if fpath.find(name)>0:
+                            enent_name = prefix[i] + name
+                            break
+         
+                    spath = os.path.join(savepath, enent_name)
+
+                    do_tracking(fpath, spath)
+        
+        k +=1
+        if k==1:
+            break
+                       
+
+
+
+def main_fold():
+    save_dir = Path('./result')
+    if not save_dir.exists():
+        save_dir.mkdir(parents=True, exist_ok=True)
+        
+    files_path = 'D:/contrast/dataset/1_to_n/709/20240709-112658_6903148351833/'   
+    for filename in os.listdir(files_path):   
+        filename = '1_track.data'
+        
+        fpath = os.path.join(files_path, filename)   
+        if os.path.isfile(fpath) and filename.find("track.data")>0:
+            # do_tracker_tracking(fpath, save_dir) 
+            do_tracking(fpath, save_dir)
+
+
+if __name__ == "__main__":
+    try:
+        main_loop()
+        # main_fold()
+    except Exception as e:
+        print(f'Error: {e}')   
+
+
+
+
+
+
+
+
+
+
+
+
+
+

tracking/rename.py

@@ -0,0 +1,35 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Sat Jun  8 09:51:59 2024
+
+@author: ym
+"""
+import os
+
+def main():
+    directory = r'D:\DetectTracking\runs\detect'  
+    directory = r'D:\DetectTracking\tracking\result\tracks'
+
+    suffix = '_'
+    
+    for root, dirs, files in os.walk(directory):    
+        for name in dirs:
+            old_name = os.path.join(root, name)
+            new_name = os.path.join(root, f"{name}{suffix}")  
+            try:
+                os.rename(old_name, new_name)
+            except Exception as e:
+                print(f"Failed to rename directory '{old_name}': {e}")
+                
+        for name in files:
+            old_name = os.path.join(root, name)
+            file, ext = os.path.splitext(name)
+            new_name = os.path.join(root, f"{file}{suffix}{ext}")
+            try:
+                os.rename(old_name, new_name)
+            except Exception as e:
+                print(f"Failed to rename file '{old_name}': {e}")
+    
+
+if __name__ == "__main__":
+    main()

tracking/test_tracking.py

@@ -12,7 +12,6 @@ import time
 import pickle
 import matplotlib.pyplot as plt
 import pandas as pd 
-
 from scipy.spatial.distance import cdist
 
 from pathlib import Path
@@ -39,8 +38,7 @@ def detect_start_end(bboxes, features_dict, filename):
     boxes = np.empty(shape=(0, 9), dtype = np.float)
     if filename.find("back") >= 0:
         vts = doBackTracks(bboxes, features_dict)
-        vtx = [t for t in vts if t.cls != 0]
+        vtx = [t for t in vts if t.cls != 0]     
-        
         for track in vtx:
             if track.moving_index.size:
                 boxes = np.concatenate((boxes, track.moving_index), axis=0)
@@ -64,7 +62,7 @@ def detect_start_end(bboxes, features_dict, filename):
         start = 0
     
     return start, end
-
+    
 def save_subimgs(vts, file, TracksDict):
     imgdir = Path(f'./result/imgs/{file}')
     if not imgdir.exists():
@@ -82,13 +80,14 @@ def save_subimgs(vts, file, TracksDict):
             cv2.imwrite(str(imgdir) + f"/{tid}_{fid}_{bid}.png", img)
 
 def have_tracked():
-    trackdict = r'./data/trackdicts'
+    trackdict = r'./data/trackdicts_20240608'
     alltracks = []
     k = 0
     gt = Profile()
     for filename in os.listdir(trackdict):
         # filename = 'test_20240402-173935_6920152400975_back_174037372.pkl'
-        # filename = '加购_91.pkl'
+        filename = '6907149227609_20240508-174733_back_returnGood_70f754088050_425_17327712807.pkl'
+        filename = '6907149227609_20240508-174733_front_returnGood_70f754088050_425_17327712807.pkl'
         
         file, ext = os.path.splitext(filename)
         filepath = os.path.join(trackdict, filename)
@@ -117,12 +116,10 @@ def have_tracked():
                 edgeline = cv2.imread("./shopcart/cart_tempt/edgeline.png")
                 draw_all_trajectories(vts, edgeline, save_dir, filename)
         print(file+f" need time: {gt.dt:.2f}s")
-        
+
-        
+        k += 1
-        
+        if k==1:
-        # k += 1
+            break
-        # if k==1:
-        #     break
     
     if len(alltracks):
         drawFeatures(alltracks, save_dir)

tracking/test_val.py

@@ -1,223 +0,0 @@
-# -*- coding: utf-8 -*-
-"""
-Created on Thu May 30 14:03:03 2024
-
-@author: ym
-"""
-import os
-import cv2
-import numpy as np
-from pathlib import Path
-import sys
-sys.path.append(r"D:\DetectTracking")
-
-
-from tracking.utils.plotting import Annotator, colors
-from tracking.utils import Boxes, IterableSimpleNamespace, yaml_load
-from tracking.trackers import BOTSORT, BYTETracker
-from tracking.dotrack.dotracks_back import doBackTracks
-from tracking.dotrack.dotracks_front import doFrontTracks 
-from tracking.utils.drawtracks import plot_frameID_y2, draw_all_trajectories
-
-W, H = 1024, 1280
-
-Mode = 'front'  #'back'
-
-def read_data_file(datapath):
-
-    with open(datapath, 'r') as file:  
-        lines = file.readlines()
-    Videos = []
-    FrameBoxes, FrameFeats = [], []
-    boxes, feats = [], []
-    
-    bboxes, ffeats = [], []
-    timestamp = []
-    t1 = None
-    for line in lines: 
-        if line.find('CameraId') >= 0: 
-            t = int(line.split(',')[1].split(':')[1])
-            timestamp.append(t)
-     
-            if len(boxes) and len(feats):
-                FrameBoxes.append(np.array(boxes, dtype = np.float32))
-                FrameFeats.append(np.array(feats, dtype = np.float32))
-                
-                boxes, feats = [], []
-            
-            if t1 and t - t1 > 1e4:
-                Videos.append((FrameBoxes, FrameFeats))    
-                FrameBoxes, FrameFeats = [], []
-            t1 = int(line.split(',')[1].split(':')[1])
-            
-        if line.find('box') >= 0:
-            box = line.split(':', )[1].split(',')[:-1]
-            boxes.append(box)
-            bboxes.append(boxes)
-            
-        
-        if line.find('feat') >= 0:
-            feat = line.split(':', )[1].split(',')[:-1]
-            feats.append(feat)
-            ffeats.append(feat)
-    
-    
-    
-    
-    FrameBoxes.append(np.array(boxes, dtype = np.float32))
-    FrameFeats.append(np.array(feats, dtype = np.float32))  
-    Videos.append((FrameBoxes, FrameFeats)) 
-    
-    TimeStamp = np.array(timestamp, dtype = np.float32) 
-    DimesDiff = np.diff((timestamp))      
-
-    return Videos
-
-def video2imgs(path):
-    vpath = os.path.join(path, "videos")
-    
-    k = 0
-    have = False
-    for filename in os.listdir(vpath):
-        file, ext = os.path.splitext(filename)
-        imgdir = os.path.join(path, file)
-        if os.path.exists(imgdir): 
-            continue          
-        else:
-            os.mkdir(imgdir)
-
-        vfile = os.path.join(vpath, filename)
-        cap = cv2.VideoCapture(vfile)
-        i = 0
-        while True:
-            ret, frame = cap.read()
-            if not ret:
-                break
-            
-            i += 1
-            imgp = os.path.join(imgdir, file+f"_{i}.png")
-            cv2.imwrite(imgp, frame)
-        
-            print(filename+f": {i}")
-        
-        
-        cap.release()
-        
-        k+=1
-        if k==1000:
-            break
-
-def draw_boxes():
-    datapath = r'D:\datasets\ym\videos_test\20240530\1_tracker_inout(1).data'
-    VideosData = read_data_file(datapath)
-    
-    bboxes = VideosData[0][0]
-    ffeats = VideosData[0][1]
-    
-    videopath = r"D:\datasets\ym\videos_test\20240530\134458234-1cd970cf-f8b9-4e80-9c2e-7ca3eec83b81-1_seek0.10415589124891511.mp4"
-    
-    cap = cv2.VideoCapture(videopath)
-    i = 0
-    while True:
-        ret, frame = cap.read()
-        if not ret:
-            break
-        
-        
-        annotator = Annotator(frame.copy(), line_width=3)
-        
-        
-        boxes = bboxes[i]
-            
-        for *xyxy, conf, cls in reversed(boxes):
-            label = f'{int(cls)}: {conf:.2f}'                   
-   
-            color = colors(int(cls), True)
-            annotator.box_label(xyxy, label, color=color)
-            
-        img = annotator.result()    
-    
-        imgpath = r"D:\datasets\ym\videos_test\20240530\result\int8_front\{}.png".format(i+1)
-        cv2.imwrite(imgpath, img)
-    
-        print(f"Output: {i}")
-        i += 1
-    cap.release()
-
-def init_tracker(tracker_yaml = None, bs=1):
-    """
-    Initialize tracker for object tracking during prediction.
-    """  
-    TRACKER_MAP = {'bytetrack': BYTETracker, 'botsort': BOTSORT}   
-    cfg = IterableSimpleNamespace(**yaml_load(tracker_yaml))
-
-    tracker = TRACKER_MAP[cfg.tracker_type](args=cfg, frame_rate=30)
-
-    return tracker
-
-def tracking(bboxes, ffeats):
-    tracker_yaml = r"./trackers/cfg/botsort.yaml"
-    tracker = init_tracker(tracker_yaml)
-    
-    track_boxes = np.empty((0, 9), dtype = np.float32)
-    features_dict = {}
-    
-    '''==================== 执行跟踪处理 ======================='''
-    for dets, feats in zip(bboxes, ffeats):
-        # 需要根据frame_id重排序
-        det_tracking = Boxes(dets).cpu().numpy()
-        tracks = tracker.update(det_tracking, feats)
-        
-        if len(tracks):
-            track_boxes = np.concatenate([track_boxes, tracks], axis=0)   
-            feat_dict = {int(x.idx): x.curr_feat for x in tracker.tracked_stracks if x.is_activated}
-            frame_id = tracks[0, 7]
-            features_dict.update({int(frame_id): feat_dict})
-    
-    return det_tracking, features_dict
-
-
-
-def main():
-    datapath = r'D:\datasets\ym\videos_test\20240530\1_tracker_inout(1).data'
-    VideosData = read_data_file(datapath)
-    
-    bboxes = VideosData[0][0]
-    ffeats = VideosData[0][1]
-    
-    bboxes, feats_dict = tracking(bboxes, ffeats)
-    
-    
-    
-    if Mode == "front":
-        vts = doFrontTracks(bboxes, feats_dict)
-        vts.classify()
-        
-
-        plt = plot_frameID_y2(vts)
-        plt.savefig('front_y2.png')  
-        # plt.close()
-    else:
-        vts = doBackTracks(bboxes, feats_dict)
-        vts.classify()
-
-        edgeline = cv2.imread("./shopcart/cart_tempt/edgeline.png")
-        draw_all_trajectories(vts, edgeline, save_dir, filename)
-    
-    
-    
-
-
-
-    
-
-
-if __name__ == "__main__":
-    filename = 'traj.png'
-    save_dir = Path('./result')
-    if not save_dir.exists():
-        save_dir.mkdir(parents=True, exist_ok=True)
-        
-    main()  
-
-

tracking/trackers/bot_sort.py

@@ -119,12 +119,14 @@ class BOTSORT(BYTETracker):
         """Returns an instance of KalmanFilterXYWH for object tracking."""
         return KalmanFilterXYWH()
 
-    def init_track(self, dets, scores, cls, imgs):
+    def init_track(self, dets, scores, cls, imgs, features_keep):
         """Initialize track with detections, scores, and classes."""
         if len(dets) == 0:
             return []
         if self.args.with_reid and self.encoder is not None:
-            features_keep = self.encoder.inference(imgs, dets)
+            if features_keep is None:
+                features_keep = self.encoder.inference(imgs, dets)
+                
             return [BOTrack(xyxy, s, c, f) for (xyxy, s, c, f) in zip(dets, scores, cls, features_keep)]  # detections
         else:
             return [BOTrack(xyxy, s, c) for (xyxy, s, c) in zip(dets, scores, cls)]  # detections

tracking/trackers/byte_tracker.py

@@ -18,7 +18,12 @@ def dists_update(dists, strack_pool, detections):
         blabel = np.array([int(stack.cls) for stack in detections])
         amlabel = np.expand_dims(alabel, axis=1).repeat(len(detections),axis=1)
         bmlabel = np.expand_dims(blabel, axis=0).repeat(len(strack_pool),axis=0)
-        dist_label = 1 - (bmlabel == amlabel)
+
+        mlabel = bmlabel == amlabel
+        iou_dist = matching.iou_distance(strack_pool, detections) > 0.1  #boxes iou>0.9时，可以不考虑类别
+        dist_label = (1 - mlabel) & iou_dist   # 不同类，且不是严格重叠，需考虑类别距离
+        
+        dist_label = 1 - mlabel
         dists = np.where(dists > dist_label, dists, dist_label)
     return dists
 
@@ -103,6 +108,7 @@ class STrack(BaseTrack):
         self.tracklet_len = 0
         self.state = TrackState.Tracked
         self.is_activated = True
+        self.first_find = False
         self.frame_id = frame_id
         if new_id:
             self.track_id = self.next_id()
@@ -127,6 +133,7 @@ class STrack(BaseTrack):
                                                                self.convert_coords(new_tlwh))
         self.state = TrackState.Tracked
         self.is_activated = True
+        self.first_find = False
 
         self.score = new_track.score
         self.cls = new_track.cls
@@ -207,7 +214,7 @@ class BYTETracker:
         self.args.new_track_thresh = 0.5
         
 
-    def update(self, results, img=None):
+    def update(self, results, img=None, features=None):
         """Updates object tracker with new detections and returns tracked object bounding boxes."""
         self.frame_id += 1
         activated_stracks = []
@@ -240,7 +247,7 @@ class BYTETracker:
         cls_keep = cls[remain_inds]
         cls_second = cls[inds_second]
 
-        detections = self.init_track(dets, scores_keep, cls_keep, img)
+        detections = self.init_track(dets, scores_keep, cls_keep, img, features)
         
         # Add newly detected tracklets to tracked_stracks
         unconfirmed = []
@@ -283,7 +290,7 @@ class BYTETracker:
         
         # Step 3: Second association, with low score detection boxes
         # association the untrack to the low score detections
-        detections_second = self.init_track(dets_second, scores_second, cls_second, img)
+        detections_second = self.init_track(dets_second, scores_second, cls_second, img, features)
         r_tracked_stracks = [strack_pool[i] for i in u_track if strack_pool[i].state == TrackState.Tracked]
         
         # TODO
@@ -366,7 +373,7 @@ class BYTETracker:
         output2 = [x.tlwh_to_tlbr(x._tlwh).tolist() + [x.track_id, x.score, x.cls, x.frame_id, x.idx] 
                               for x in first_finded if x.first_find]
         
-        output = np.asarray(output1+output2, dtype=np.float32)
+        output = np.asarray(output1 + output2, dtype=np.float32)
 
         return output
     
@@ -382,7 +389,7 @@ class BYTETracker:
         tracks = []
         feats = []                  
         for t in self.tracked_stracks:
-            if t.is_activated:
+            if t.is_activated or t.first_find:
                 track = t.tlbr.tolist() + [t.track_id, t.score, t.cls, t.idx] 
                 feat = t.curr_feature
                 
@@ -398,7 +405,7 @@ class BYTETracker:
         """Returns a Kalman filter object for tracking bounding boxes."""
         return KalmanFilterXYAH()
 
-    def init_track(self, dets, scores, cls, img=None):
+    def init_track(self, dets, scores, cls, img=None, feats=None):
         """Initialize object tracking with detections and scores using STrack algorithm."""
         return [STrack(xyxy, s, c) for (xyxy, s, c) in zip(dets, scores, cls)] if len(dets) else []  # detections
 
@@ -455,7 +462,22 @@ class BYTETracker:
     def remove_duplicate_stracks(stracksa, stracksb):
         """Remove duplicate stracks with non-maximum IOU distance."""
         pdist = matching.iou_distance(stracksa, stracksb)
-        pairs = np.where(pdist < 0.15)
+        
+        #### ===================================== written by WQG
+        mlabel = []
+        if len(stracksa) and len(stracksb):
+            alabel = np.array([int(stack.cls) for stack in stracksa])
+            blabel = np.array([int(stack.cls) for stack in stracksb])
+            amlabel = np.expand_dims(alabel, axis=1).repeat(len(stracksb),axis=1)
+            bmlabel = np.expand_dims(blabel, axis=0).repeat(len(stracksa),axis=0)
+            mlabel = bmlabel == amlabel
+        if len(mlabel):
+            condt = (pdist<0.15) & mlabel   # 需满足iou足够小，且类别相同，才予以排除
+        else:
+            condt = pdist<0.15
+        
+        
+        pairs = np.where(condt)
         dupa, dupb = [], []
         for p, q in zip(*pairs):
             timep = stracksa[p].frame_id - stracksa[p].start_frame

tracking/trackers/reid/reid_interface.py

@@ -45,8 +45,7 @@ class ReIDInterface:
             ])
         
         
-        self.model = nn.DataParallel(model).to(self.device)
+        # self.model = nn.DataParallel(model).to(self.device)
-        
         self.model = model
         self.model.load_state_dict(torch.load(self.model_path, map_location=self.device))
         self.model.eval()

tracking/utils/drawtracks.py

@@ -80,15 +80,32 @@ def plot_frameID_y2(vts):
     return plt
 
 
-def draw_all_trajectories(vts, edgeline, save_dir, filename):
+def draw_all_trajectories(vts, edgeline, save_dir, file, draw5p=False):
     '''显示四种类型结果'''
-    file, ext = os.path.splitext(filename)
+    # file, ext = os.path.splitext(filename)
     # edgeline = cv2.imread("./shopcart/cart_tempt/edgeline.png")
     # edgeline2 = edgeline1.copy()
     # edgeline = np.concatenate((edgeline1, edgeline2), exis = 1)
     
-    '''1. tracks 5点轨迹'''
+    if not isinstance(save_dir, Path): save_dir = Path(save_dir)
-    trackpth = save_dir.parent /Path("trajectory")/ Path(f"{file}")
+    
+    ''' all tracks 中心轨迹'''
+    img1, img2 = edgeline.copy(), edgeline.copy()
+    img1 = drawTrack(vts.tracks, img1)
+    img2 = drawTrack(vts.Residual, img2)
+    
+    img = np.concatenate((img1, img2), axis = 1)
+    H, W = img.shape[:2]
+    cv2.line(img, (int(W/2), 0), (int(W/2), H), (128, 255, 128), 2)
+    
+    imgpth = save_dir.joinpath(f"{file}_show.png")
+    cv2.imwrite(str(imgpth), img)
+
+    if not draw5p:
+        return
+        
+    ''' tracks 5点轨迹'''
+    trackpth = save_dir / Path("trajectory") / Path(f"{file}")
     if not trackpth.exists():
         trackpth.mkdir(parents=True, exist_ok=True)
     for track in vts.tracks:
@@ -106,23 +123,9 @@ def draw_all_trajectories(vts, edgeline, save_dir, filename):
 
         pth = trackpth.joinpath(f"{track.tid}_.png")        
         cv2.imwrite(str(pth), img)
+
     
     
-    
-    
-    '''2. all tracks 中心轨迹'''
-    img1, img2 = edgeline.copy(), edgeline.copy()
-    
-    img1 = drawTrack(vts.tracks, img1)
-    img2 = drawTrack(vts.Residual, img2)
-    
-    img = np.concatenate((img1, img2), axis = 1)
-    H, W = img.shape[:2]
-    cv2.line(img, (int(W/2), 0), (int(W/2), H), (128, 255, 128), 2)
-    
-    pth = save_dir.joinpath(f"{file}_show.png")
-    cv2.imwrite(str(pth), img)
-    
 # =============================================================================
 #     '''3. moving tracks 中心轨迹'''
 #     filename2 = f"{file}_show_r.png"
@@ -134,13 +137,11 @@ def draw_all_trajectories(vts, edgeline, save_dir, filename):
     
 
     
-# =============================================================================
+    '''5. tracks 时序：trajmin、trajmax、arearate、incartrate'''
-#     '''5. tracks 时序：trajmin、trajmax、arearate、incartrate'''
+    # plt = drawtracefeat(vts)
-#     plt = drawtracefeat(vts)
+    # pth = save_dir.joinpath(f"{file}_x.png")
-#     pth = save_dir.joinpath(f"{file}_x.png")
+    # plt.savefig(pth)
-#     plt.savefig(pth)
+    # plt.close('all')
-#     plt.close('all')
-# =============================================================================
     
 
 

tracking/utils/mergetrack.py

@@ -15,17 +15,18 @@ def readDict(boxes, TracksDict):
     for i in range(boxes.shape[0]):   
         tid, fid, bid = int(boxes[i, 4]), int(boxes[i, 7]), int(boxes[i, 8])
         
-        feat = TracksDict[f"frame_{fid}"]["feats"][bid]
+        trackdict = TracksDict[f"frame_{fid}"]
-        img = TracksDict[f"frame_{fid}"]["imgs"][bid]
+        if "feats" in trackdict:
+            feat = trackdict["feats"][bid]
+            feats.append(feat)
         
-        box = TracksDict[f"frame_{fid}"]["boxes"][bid]
+        if "boxes" in trackdict:
+            box = trackdict["boxes"][bid]
+            assert (box[:4].astype(int) == boxes[i, :4].astype(int)).all(), f"Please check: frame_{fid}"
         
-        assert (box[:4].astype(int) == boxes[i, :4].astype(int)).all(), f"Please check: frame_{fid}"
+        if "imgs" in trackdict:
-        
+            img = trackdict["imgs"][bid]
-        feats.append(feat)
+            cv2.imwrite(f'./data/imgs/{tid}_{fid}_{bid}.png', img)
-        
-        # img = TracksDict[fid][f'{bid}_img']
-        # cv2.imwrite(f'./data/imgs/{tid}_{fid}_{bid}.png', img)
     
     return np.asarray(feats, dtype=np.float32)
 
@@ -59,13 +60,12 @@ def track_equal_track(atrack, btrack):
     ''' 2. 轨迹特征相似度判断'''
     feat = np.concatenate((afeat, bfeat), axis=0)
     
-    emb_simil = 1-np.maximum(0.0, cdist(feat, feat, 'cosine'))
+    emb_simil = 1 - np.maximum(0.0, cdist(feat, feat, 'cosine'))
-    emb_ = 1-cdist(np.mean(afeat, axis=0)[None, :], np.mean(bfeat, axis=0)[None, :], 'cosine')
+    emb_ = 1 - np.maximum(0.0, cdist(np.mean(afeat, axis=0)[None, :], np.mean(bfeat, axis=0)[None, :], 'cosine'))/2
     
     if emb_[0, 0]<0.66:
         return False
     
-    
     ''' 3. 轨迹空间iou'''
     alabel = np.array([0] * afids.size, dtype=np.int_)
     blabel = np.array([1] * bfids.size, dtype=np.int_)
@@ -93,7 +93,7 @@ def track_equal_track(atrack, btrack):
         
         af, bf = afeat[a, :], bfeat[b, :]
         
-        emb_ab = 1-cdist(af[None, :], bf[None, :], 'cosine')
+        emb_ab = 1 - np.maximum(0.0, cdist(af[None, :], bf[None, :], 'cosine'))
         
         
         xa1, ya1 = abox[0] - abox[2]/2, abox[1] - abox[3]/2 
@@ -113,7 +113,22 @@ def track_equal_track(atrack, btrack):
 
         ious.append(inter/union)
         embs.append(emb_ab[0, 0])
-
+    
+    ''' 4. 和同一手部关联，如何将该代码和 iou 部分相融合,需进一步完善'''   
+    # ahands = np.array(atrack.Hands)
+    # bhands = np.array(btrack.Hands)
+    # ahids = ahands[:, 0]
+    # bhids = bhands[:, 0]
+    # interhid = set(ahids).intersection(set(bhids))
+    # for hid in interhid:
+    #     aidx = ahands[:, 0] == hid
+    #     bidx = bhands[:, 0] == hid
+        
+    #     ahfids = ahids[aidx, 1]
+    #     bhfids = bhids[bidx, 1]
+    
+    
+    
 
     cont = False if len(interfid) else True    # fid 无交集
     cont1 = all(emb > 0.5 for emb in embs)

tracking/utils/plotting.py

@@ -1,4 +1,4 @@
-# Ultralytics YOLO 🚀, AGPL-3.0 license
+# Ultralytics YOLO 🚀, AGPL-3.0 license
 
 import contextlib
 import math
@@ -284,5 +284,59 @@ def boxing_img(det, img, line_width=3):
     imgx = annotator.result()
     
     return imgx
+
+def draw_tracking_boxes(imgs, tracks, scale=2):
+    '''tracks: [x1, y1, x2, y2, track_id, score, cls, frame_index, box_index]
+                0   1   2   3      4        5     6       7           8
+       关键：
+         (1) imgs中的次序和 track 中的 fid 对应
+         (2) img 尺度小对于xyxy减半
+       
+    '''
+    
+    def array2list(bboxes):
+        track_fids = np.unique(bboxes[:, 7].astype(int))
+        track_fids.sort()
+        
+        lboxes = []
+        for f_id in track_fids:
+            # print(f"The ID is: {t_id}")
+            idx = np.where(bboxes[:, 7] == f_id)[0]
+            box = bboxes[idx, :]  
+            lboxes.append(box)
+            
+            assert len(set(box[:, 4])) == len(box), "Please check!!!"
+            
+        return lboxes
+    
+    bboxes = array2list(tracks)
+    
+    if len(bboxes)!=len(imgs):
+        return []
+    
+    subimgs = []
+    for i, boxes in enumerate(bboxes):
+        annotator = Annotator(imgs[i].copy())
+        for *xyxy, tid, conf, cls, fid, bid in boxes:
+            label = f'id:{int(tid)}_{int(cls)}_{conf:.2f}'
+            
+            if cls==0:
+                color = colors(int(cls), True)
+            elif tid>0 and cls!=0:
+                color = colors(int(tid), True)
+            else:
+                color = colors(19, True)     # 19为调色板的最后一个元素
+            
+            pt2 = [p/scale for p in xyxy]
+            annotator.box_label(pt2, label, color=color)
+        
+        img = annotator.result()
+        subimgs.append(img)
+        
+    return subimgs
+
+
+
+
     
 

tracking/utils/proBoxes.py

@@ -12,10 +12,12 @@ class Boxes:
         """Initialize the Boxes class."""
         if boxes.ndim == 1:
             boxes = boxes[None, :]
-        n = boxes.shape[-1]
+        m, n = boxes.shape
-        assert n in (6, 7, 8), f'expected `n` in [6, 7], but got {n}'  # xyxyb, track_id, conf, cls
+        assert n in (6, 7), f'expected `n` in [6, 7], but got {n}'  # xyxy, track_id, conf, cls
+        
+        '''对每一个box进行编号，利用该编号可以索引对应 feature'''
+        self.data = np.concatenate([boxes[:, :4], np.arange(m).reshape(-1, 1), boxes[:, 4:]], axis=-1)
         
-        self.data = boxes
         self.orig_shape = orig_shape
     
     def cpu(self):
@@ -30,10 +32,9 @@ class Boxes:
         """Return the boxes in xyxy format."""
         return self.data[:, :4]
     
-    
     @property
     def xyxyb(self):
-        """Return the boxes in xyxyb format."""
+        """Return the boxes in xyxyb format."""    
         return self.data[:, :5]
 
     @property

tracking/utils/read_data.py

@@ -0,0 +1,236 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Fri Jul  5 13:59:21 2024
+func: extract_data()
+  读取 Pipeline 各模块的数据，在 read_pipeline_data.py（马晓慧）的基础上完成接口改造
+
+@author: ym
+"""
+import numpy as np
+import re
+import os
+
+
+
+
+def str_to_float_arr(s):
+    # 移除字符串末尾的逗号（如果存在）  
+    if s.endswith(','):  
+        s = s[:-1]  
+      
+    # 使用split()方法分割字符串，然后将每个元素转化为float  
+    float_array = [float(x) for x in s.split(",")]
+    return float_array        
+
+
+def find_samebox_in_array(arr, target):
+
+    for i, st in enumerate(arr):
+        if st[:4] == target[:4]:
+            return i
+    return -1
+
+
+def extract_data(datapath):
+    bboxes, ffeats = [], []
+    
+    trackerboxes = np.empty((0, 9), dtype=np.float64)
+    trackerfeats = np.empty((0, 256), dtype=np.float64)
+    
+    boxes, feats, tboxes, tfeats = [], [], [], []
+    with open(datapath, 'r', encoding='utf-8') as lines:
+        for line in lines:  
+            line = line.strip()  # 去除行尾的换行符和可能的空白字符  
+            if not line:  # 跳过空行  
+                continue
+            
+            if line.find("CameraId")>=0:
+                if len(boxes): bboxes.append(np.array(boxes))
+                if len(feats): ffeats.append(np.array(feats))
+                if len(tboxes):  
+                    trackerboxes = np.concatenate((trackerboxes, np.array(tboxes)))
+                if len(tfeats):
+                    trackerfeats = np.concatenate((trackerfeats, np.array(tfeats)))
+     
+                boxes, feats, tboxes, tfeats = [], [], [], []
+            
+            if line.find("box:") >= 0 and line.find("output_box:") < 0:
+                box = line[line.find("box:") + 4:].strip()
+                boxes.append(str_to_float_arr(box))
+                
+            if line.find("feat:") >= 0:
+                feat = line[line.find("feat:") + 5:].strip()
+                feats.append(str_to_float_arr(feat))
+                    
+            if line.find("output_box:") >= 0:
+                box = str_to_float_arr(line[line.find("output_box:") + 11:].strip())
+                tboxes.append(box)  # 去掉'output_box:'并去除可能的空白字符  
+                index = find_samebox_in_array(boxes, box)
+                if index >= 0:
+                    # feat_f = str_to_float_arr(input_feats[index])
+                    feat_f = feats[index]
+                    norm_f = np.linalg.norm(feat_f) 
+                    feat_f = feat_f / norm_f
+                    tfeats.append(feat_f)
+    
+    if len(boxes): bboxes.append(np.array(boxes))
+    if len(feats): ffeats.append(np.array(feats))
+    if len(tboxes): trackerboxes = np.concatenate((trackerboxes, np.array(tboxes)))
+    if len(tfeats): trackerfeats = np.concatenate((trackerfeats, np.array(tfeats)))
+    
+    assert(len(bboxes)==len(ffeats)), "Error at Yolo output!"
+    assert(len(trackerboxes)==len(trackerfeats)), "Error at tracker output!"
+    
+    tracker_feat_dict = {}
+    for i in range(len(trackerboxes)):
+        tid, fid, bid = int(trackerboxes[i, 4]), int(trackerboxes[i, 7]), int(trackerboxes[i, 8])
+        if f"frame_{fid}" not in tracker_feat_dict:
+            tracker_feat_dict[f"frame_{fid}"]= {"feats": {}}
+        tracker_feat_dict[f"frame_{fid}"]["feats"].update({bid: trackerfeats[i, :]})
+    
+    
+    boxes, trackingboxes= [], []
+    tracking_flag = False    
+    with open(datapath, 'r', encoding='utf-8') as lines:
+        for line in lines:  
+            line = line.strip()  # 去除行尾的换行符和可能的空白字符  
+            if not line:  # 跳过空行  
+                continue       
+            if tracking_flag:
+                if line.find("tracking_") >= 0:
+                    tracking_flag = False
+                else:
+                    box = str_to_float_arr(line)
+                    boxes.append(box)
+            if line.find("tracking_") >= 0:
+                tracking_flag = True
+                if len(boxes):
+                    trackingboxes.append(np.array(boxes))
+                boxes = []
+    
+    if len(boxes): 
+        trackingboxes.append(np.array(boxes))
+    
+    tracking_feat_dict = {}
+    for i, boxes in enumerate(trackingboxes):
+        for box in boxes:
+            tid, fid, bid = int(box[4]), int(box[7]), int(box[8])
+            if f"track_{tid}" not in tracking_feat_dict:
+                tracking_feat_dict[f"track_{tid}"]= {"feats": {}}
+            tracking_feat_dict[f"track_{tid}"]["feats"].update({f"{fid}_{bid}": tracker_feat_dict[f"frame_{fid}"]["feats"][bid]})
+
+    return bboxes, ffeats, trackerboxes, tracker_feat_dict, trackingboxes, tracking_feat_dict
+
+def read_tracking_output(filepath):
+    boxes = []
+    feats = []
+    with open(filepath, 'r', encoding='utf-8') as file:
+        for line in file:  
+            line = line.strip()  # 去除行尾的换行符和可能的空白字符
+            
+            if not line:  
+                continue 
+            
+            if line.endswith(','):  
+                line = line[:-1]
+                
+            data = np.array([float(x) for x in line.split(",")])
+            if data.size == 9:
+                boxes.append(data)
+            if data.size == 256:
+                feats.append(data)
+    
+    return np.array(boxes), np.array(feats) 
+    
+
+def read_deletedBarcode_file(filePth):
+    with open(filePth, 'r', encoding='utf-8') as f:
+        lines = f.readlines()
+
+        split_flag, all_list = False, []
+        dict, barcode_list, similarity_list =  {}, [], []
+  
+        clean_lines = [line.strip().replace("'", '').replace('"', '') for line in lines]
+        for line in clean_lines:
+            stripped_line = line.strip()
+            if not stripped_line:
+                if len(barcode_list): dict['barcode'] = barcode_list
+                if len(similarity_list): dict['similarity'] = similarity_list
+                if len(dict): all_list.append(dict)
+                
+                split_flag = False
+                dict, barcode_list, similarity_list =  {}, [], []
+                continue
+            
+            # print(line)
+            label = line.split(':')[0]
+            value = line.split(':')[1]
+            if label == 'SeqDir':
+                dict['SeqDir'] = value
+            if label == 'Deleted':
+                dict['Deleted'] = value
+            if label == 'List':
+                split_flag = True
+                continue
+            if split_flag:
+                barcode_list.append(label)
+                similarity_list.append(value)
+        
+        if len(barcode_list): dict['barcode'] = barcode_list
+        if len(similarity_list): dict['similarity'] = similarity_list
+        if len(dict): all_list.append(dict)
+    return all_list
+
+
+
+
+if __name__ == "__main__":    
+    files_path = 'D:/contrast/dataset/1_to_n/709/20240709-112658_6903148351833/'  
+    
+    # 遍历目录下的所有文件和目录  
+    for filename in os.listdir(files_path):    
+        filename = '1_track.data'
+        file_path = os.path.join(files_path, filename)
+        if os.path.isfile(file_path) and filename.find("track.data")>0:
+            extract_data(file_path)
+ 
+            print("Done")
+
+    
+    
+    
+    
+    
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

tracking/utils/read_pipeline_data.py

@@ -0,0 +1,250 @@
+# -*- coding: utf-8 -*-
+"""
+Created on Tue May 21 15:25:23 2024
+读取 Pipeline 各模块的数据，主代码由 马晓慧 完成
+
+@author: ieemoo-zl003
+"""
+
+import os
+import numpy as np
+  
+# 替换为你的目录路径  
+files_path = 'D:/contrast/dataset/1_to_n/709/20240709-112658_6903148351833/'  
+
+def str_to_float_arr(s):
+    # 移除字符串末尾的逗号（如果存在）  
+    if s.endswith(','):  
+        s = s[:-1]  
+      
+    # 使用split()方法分割字符串，然后将每个元素转化为float  
+    float_array = np.array([float(x) for x in s.split(",")]) 
+    return float_array
+
+def extract_tracker_input_boxes_feats(file_name):
+    boxes = []
+    feats = []
+    with open(file_name, 'r', encoding='utf-8') as file:
+        for line in file:  
+            line = line.strip()  # 去除行尾的换行符和可能的空白字符  
+              
+            # 跳过空行  
+            if not line:  
+                continue  
+              
+            # 检查是否以'box:'或'feat:'开始  
+            if line.find("box:") >= 0 and line.find("output_box:") < 0:  
+                box = line[line.find("box:") + 4:].strip()
+                boxes.append(str_to_float_arr(box))  # 去掉'box:'并去除可能的空白字符  
+
+            if line.find("feat:") >= 0:
+                feat = line[line.find("feat:") + 5:].strip()
+                feats.append(str_to_float_arr(feat))  # 去掉'box:'并去除可能的空白字符 
+            
+    return np.array(boxes), np.array(feats)
+
+def find_string_in_array(arr, target):  
+    """  
+    在字符串数组中找到目标字符串对应的行（索引）。  
+  
+    参数:  
+    arr -- 字符串数组  
+    target -- 要查找的目标字符串  
+  
+    返回:  
+    目标字符串在数组中的索引。如果未找到，则返回-1。  
+    """  
+    tg = [float(t) for k, t in enumerate(target.split(',')) if k<4][:4]
+    for i, st in enumerate(arr): 
+        st = [float(s) for k, s in enumerate(target.split(',')) if k<4][:4]
+        
+        if st == tg:
+            return i
+        
+        # if st[:20] == target[:20]:  
+        #     return i  
+    return -1 
+
+def find_samebox_in_array(arr, target):
+
+    for i, st in enumerate(arr):
+        if all(st[:4] == target[:4]):
+            return i
+    return -1
+    
+    
+    
+
+
+
+def extract_tracker_output_boxes_feats(read_file_name):
+    
+    input_boxes, input_feats = extract_tracker_input_boxes_feats(read_file_name)
+    
+    boxes = []
+    feats = []
+    with open(read_file_name, 'r', encoding='utf-8') as file:
+        for line in file:  
+            line = line.strip()  # 去除行尾的换行符和可能的空白字符  
+              
+            # 跳过空行  
+            if not line:  
+                continue  
+              
+            # 检查是否以'output_box:'开始  
+            if line.find("output_box:") >= 0:
+                box = str_to_float_arr(line[line.find("output_box:") + 11:].strip())
+                boxes.append(box)  # 去掉'output_box:'并去除可能的空白字符  
+                index = find_samebox_in_array(input_boxes, box)
+                if index >= 0:
+                    # feat_f = str_to_float_arr(input_feats[index])
+                    feat_f = input_feats[index]
+                    norm_f = np.linalg.norm(feat_f) 
+                    feat_f = feat_f / norm_f
+                    feats.append(feat_f)
+    return input_boxes, input_feats, np.array(boxes), np.array(feats)
+
+def extract_tracking_output_boxes_feats(read_file_name):
+    tracker_boxes, tracker_feats, input_boxes, input_feats = extract_tracker_output_boxes_feats(read_file_name)
+    boxes = []
+    feats = []
+    
+    tracking_flag = False
+    with open(read_file_name, 'r', encoding='utf-8') as file:
+        for line in file:  
+            line = line.strip()  # 去除行尾的换行符和可能的空白字符  
+              
+            # 跳过空行  
+            if not line:  
+                continue  
+            
+            if tracking_flag:
+                if line.find("tracking_") >= 0:
+                    tracking_flag = False
+                else:
+                    box = str_to_float_arr(line)
+                    boxes.append(box)
+                    index = find_samebox_in_array(input_boxes, box)
+                    if index >= 0:
+                        feats.append(input_feats[index])
+            # 检查是否以tracking_'开始  
+            if line.find("tracking_") >= 0:
+                tracking_flag = True
+
+    assert(len(tracker_boxes)==len(tracker_feats)), "Error at Yolo output"
+    assert(len(input_boxes)==len(input_feats)), "Error at tracker output"
+    assert(len(boxes)==len(feats)), "Error at tracking output"            
+
+    return tracker_boxes, tracker_feats, input_boxes, input_feats, np.array(boxes), np.array(feats)
+
+def read_tracking_input(datapath):
+    with open(datapath, 'r') as file:  
+        lines = file.readlines()
+    
+    data = []
+    for line in lines:
+        data.append([s for s in line.split(',') if len(s)>=3])
+        # data.append([float(s) for s in line.split(',') if len(s)>=3])
+
+    # data = np.array(data, dtype = np.float32)
+    try:    
+        data = np.array(data, dtype = np.float32)
+    except Exception as e:
+        data = np.array([], dtype = np.float32)
+        print('DataError for func: read_tracking_input()')
+        
+    
+    return data
+
+
+
+def read_tracker_input(datapath):
+    with open(datapath, 'r') as file:  
+        lines = file.readlines()
+    Videos = []
+    FrameBoxes, FrameFeats = [], []
+    boxes, feats = [], []
+    
+    timestamp = []
+    t1 = None
+    for line in lines: 
+        if line.find('CameraId') >= 0: 
+            t = int(line.split(',')[1].split(':')[1])
+            timestamp.append(t)
+     
+            if len(boxes) and len(feats):
+                FrameBoxes.append(np.array(boxes, dtype = np.float32))
+                FrameFeats.append(np.array(feats, dtype = np.float32))
+                boxes, feats = [], []
+            
+            if t1 and t - t1 > 1e3:
+                Videos.append((FrameBoxes, FrameFeats))    
+                FrameBoxes, FrameFeats = [], []
+            t1 = int(line.split(',')[1].split(':')[1])
+            
+        if line.find('box') >= 0:
+            box = line.split(':', )[1].split(',')[:-1]
+            boxes.append(box)
+            
+        
+        if line.find('feat') >= 0:
+            feat = line.split(':', )[1].split(',')[:-1]
+            feats.append(feat)
+
+    FrameBoxes.append(np.array(boxes, dtype = np.float32))
+    FrameFeats.append(np.array(feats, dtype = np.float32))  
+    Videos.append((FrameBoxes, FrameFeats)) 
+    
+    # TimeStamp = np.array(timestamp, dtype = np.int64) 
+    # DimesDiff = np.diff((TimeStamp))
+    # sorted_indices = np.argsort(TimeStamp)
+    # TimeStamp_sorted = TimeStamp[sorted_indices]
+    # DimesDiff_sorted = np.diff((TimeStamp_sorted))
+    
+    return Videos
+
+
+
+def main():
+    files_path = 'D:/contrast/dataset/1_to_n/709/20240709-112658_6903148351833/'  
+    
+    # 遍历目录下的所有文件和目录  
+    for filename in os.listdir(files_path):  
+        # 构造完整的文件路径  
+        file_path = os.path.join(files_path, filename)
+        if os.path.isfile(file_path) and filename.find("track.data")>0:
+            tracker_boxes, tracker_feats, tracking_boxes, tracking_feats, output_boxes, output_feats = extract_tracking_output_boxes_feats(file_path)
+ 
+            print("Done")
+            
+        
+if __name__ == "__main__":
+    main()     
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        
+        

videos_select.py

@@ -0,0 +1,641 @@
+# YOLOv5 🚀 by Ultralytics, AGPL-3.0 license
+"""
+Run YOLOv5 detection inference on images, videos, directories, globs, YouTube, webcam, streams, etc.
+
+Usage - sources:
+    $ python detect.py --weights yolov5s.pt --source 0                               # webcam
+                                                     img.jpg                         # image
+                                                     vid.mp4                         # video
+                                                     screen                          # screenshot
+                                                     path/                           # directory
+                                                     list.txt                        # list of images
+                                                     list.streams                    # list of streams
+                                                     'path/*.jpg'                    # glob
+                                                     'https://youtu.be/Zgi9g1ksQHc'  # YouTube
+                                                     'rtsp://example.com/media.mp4'  # RTSP, RTMP, HTTP stream
+
+Usage - formats:
+    $ python detect.py --weights yolov5s.pt                 # PyTorch
+                                 yolov5s.torchscript        # TorchScript
+                                 yolov5s.onnx               # ONNX Runtime or OpenCV DNN with --dnn
+                                 yolov5s_openvino_model     # OpenVINO
+                                 yolov5s.engine             # TensorRT
+                                 yolov5s.mlmodel            # CoreML (macOS-only)
+                                 yolov5s_saved_model        # TensorFlow SavedModel
+                                 yolov5s.pb                 # TensorFlow GraphDef
+                                 yolov5s.tflite             # TensorFlow Lite
+                                 yolov5s_edgetpu.tflite     # TensorFlow Edge TPU
+                                 yolov5s_paddle_model       # PaddlePaddle
+"""
+
+import argparse
+import csv
+import os
+import platform
+import sys
+from pathlib import Path
+import glob
+import numpy as np
+import pickle
+import torch
+
+FILE = Path(__file__).resolve()
+ROOT = FILE.parents[0]  # YOLOv5 root directory
+if str(ROOT) not in sys.path:
+    sys.path.append(str(ROOT))  # add ROOT to PATH
+sys.path.append((str(ROOT)+'\\tracking\\utils'))
+
+ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative
+
+
+from models.common import DetectMultiBackend
+from utils.dataloaders import IMG_FORMATS, VID_FORMATS, LoadImages, LoadScreenshots, LoadStreams
+from utils.general import (LOGGER, Profile, check_file, check_img_size, check_imshow, check_requirements, colorstr, cv2,
+                           increment_path, non_max_suppression, print_args, scale_boxes, strip_optimizer, xyxy2xywh)
+from utils.torch_utils import select_device, smart_inference_mode
+
+'''集成跟踪模块，输出跟踪结果文件 .npy'''
+# from ultralytics.engine.results import Boxes    # Results
+# from ultralytics.utils import IterableSimpleNamespace, yaml_load
+from tracking.utils.plotting import Annotator, colors
+from tracking.utils import Boxes, IterableSimpleNamespace, yaml_load, boxes_add_fid
+from tracking.trackers import BOTSORT, BYTETracker
+from tracking.utils.showtrack import drawtracks
+from hands.hand_inference import hand_pose
+from tracking.dotrack.dotracks_back import doBackTracks
+from tracking.dotrack.dotracks_front import doFrontTracks
+
+from tracking.trackers.reid.reid_interface import ReIDInterface 
+from tracking.trackers.reid.config import config as ReIDConfig
+ReIDEncoder = ReIDInterface(ReIDConfig)
+
+# tracker_yaml = r"./tracking/trackers/cfg/botsort.yaml"
+
+def inference_image(image, detections):
+    H, W, _ = np.shape(image)
+    imgs = []
+    batch_patches = []
+    patches = []
+    for d in range(np.size(detections, 0)):
+        tlbr = detections[d, :4].astype(np.int_)
+        tlbr[0] = max(0, tlbr[0])
+        tlbr[1] = max(0, tlbr[1])
+        tlbr[2] = min(W - 1, tlbr[2])
+        tlbr[3] = min(H - 1, tlbr[3])
+        img1 = image[tlbr[1]:tlbr[3], tlbr[0]:tlbr[2], :]
+
+        img = img1[:, :, ::-1].copy()    # the model expects RGB inputs
+        patch = ReIDEncoder.transform(img)
+        
+        imgs.append(img1)
+        # 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 imgs, features
+
+
+
+def init_trackers(tracker_yaml = None, bs=1):
+    """
+    Initialize trackers for object tracking during prediction.
+    """      
+    # tracker_yaml = r"./tracking/trackers/cfg/botsort.yaml"
+    
+    TRACKER_MAP = {'bytetrack': BYTETracker, 'botsort': BOTSORT}
+    
+    cfg = IterableSimpleNamespace(**yaml_load(tracker_yaml))
+    trackers = []
+    for _ in range(bs):
+        tracker = TRACKER_MAP[cfg.tracker_type](args=cfg, frame_rate=30)
+        trackers.append(tracker)
+    
+    return trackers
+
+def detect_start_end(bboxes, features_dict, filename):
+    
+    boxes = np.empty(shape=(0, 9), dtype = np.float64)
+    if filename.find("back") >= 0:
+        vts = doBackTracks(bboxes, features_dict)
+        vts.classify()
+        # vtx = [t for t in vts.tracks if t.cls != 0]
+        # for track in vtx:
+        for track in vts.Residual:
+            if track.moving_index.size:
+                boxes = np.concatenate((boxes, track.moving_index), axis=0)
+                
+    elif filename.find("front") >= 0:
+        vts = doFrontTracks(bboxes, features_dict)
+        vts.classify()
+        # vtx = [t for t in vts.tracks if t.cls != 0]
+        # for track in vtx:
+        for track in vts.Residual:
+            for start, end in track.dynamic_y2:
+                boxes = np.concatenate((boxes, track.boxes[start:end+1, :]), axis=0)    
+            for start, end in track.dynamic_y1:
+                boxes = np.concatenate((boxes, track.boxes[start:end+1, :]), axis=0)
+    
+    if boxes.size > 0:
+        start = np.min(boxes[:, 7])
+        end = np.max(boxes[:, 7])
+    else:
+        start, end = 0, 0
+    
+    return start, end
+
+@smart_inference_mode()
+def run(
+        weights=ROOT / 'yolov5s.pt',  # model path or triton URL
+        source=ROOT / 'data/images',  # file/dir/URL/glob/screen/0(webcam)
+        
+        project=ROOT / 'runs/detect',  # save results to project/name
+        name='exp',  # save results to project/name
+        
+        tracker_yaml = "./tracking/trackers/cfg/botsort.yaml",
+        imgsz=(640, 640),  # inference size (height, width)
+        conf_thres=0.25,  # confidence threshold
+        iou_thres=0.45,  # NMS IOU threshold
+        max_det=1000,  # maximum detections per image
+        device='',  # cuda device, i.e. 0 or 0,1,2,3 or cpu
+        view_img=False,  # show results
+        save_txt=False,  # save results to *.txt
+        save_csv=False,  # save results in CSV format
+        save_conf=False,  # save confidences in --save-txt labels
+        save_crop=False,  # save cropped prediction boxes
+        nosave=False,  # do not save images/videos
+        classes=None,  # filter by class: --class 0, or --class 0 2 3
+        agnostic_nms=False,  # class-agnostic NMS
+        augment=False,  # augmented inference
+        visualize=False,  # visualize features
+        update=False,  # update all models
+        exist_ok=False,  # existing project/name ok, do not increment
+        line_thickness=3,  # bounding box thickness (pixels)
+        hide_labels=False,  # hide labels
+        hide_conf=False,  # hide confidencesL
+        half=False,  # use FP16 half-precision inference
+        dnn=False,  # use OpenCV DNN for ONNX inference
+        vid_stride=1,  # video frame-rate stride
+        data=ROOT / 'data/coco128.yaml',  # dataset.yaml path
+):
+    source = str(source)
+    # filename = os.path.split(source)[-1]
+    
+    save_img = not nosave and not source.endswith('.txt')  # save inference images
+    is_file = Path(source).suffix[1:] in (IMG_FORMATS + VID_FORMATS)
+    is_url = source.lower().startswith(('rtsp://', 'rtmp://', 'http://', 'https://'))
+    webcam = source.isnumeric() or source.endswith('.streams') or (is_url and not is_file)
+    screenshot = source.lower().startswith('screen')
+    if is_url and is_file:
+        source = check_file(source)  # download
+    
+    save_dir = Path(project) / Path(source).stem
+    if save_dir.exists():
+        print(Path(source).stem)
+        # return
+        
+        save_dir = increment_path(Path(project) / name, exist_ok=exist_ok)  # increment run
+        (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True)  # make dir
+    else:
+        save_dir.mkdir(parents=True, exist_ok=True) 
+        
+    # Load model
+    device = select_device(device)
+    model = DetectMultiBackend(weights, device=device, dnn=dnn, data=data, fp16=half)
+    stride, names, pt = model.stride, model.names, model.pt
+    imgsz = check_img_size(imgsz, s=stride)  # check image size
+     
+    # Dataloader
+    bs = 1  # batch_size
+    
+    dataset = LoadImages(source, img_size=imgsz, stride=stride, auto=pt, vid_stride=vid_stride)    
+    vid_path, vid_writer = [None] * bs, [None] * bs
+   
+    
+    # Run inference
+    model.warmup(imgsz=(1 if pt or model.triton else bs, 3, *imgsz))  # warmup
+    seen, dt = 0, (Profile(), Profile(), Profile())
+    
+    tracker = init_trackers(tracker_yaml, bs)[0]
+    
+    handpose = hand_pose()
+    handlocals_dict = {}
+    
+    boxes_and_imgs  = []
+    
+    BoxesFeats = []
+    
+    
+    
+    track_boxes = np.empty((0, 9), dtype = np.float32)
+    det_boxes = np.empty((0, 9), dtype = np.float32)
+    
+    DetBoxes = np.empty((0, 6), dtype = np.float32)
+    TrackerBoxes = np.empty((0, 9), dtype = np.float32)
+    TrackerFeats = np.empty((0, 256), dtype = np.float32)
+    
+    images = []
+    
+    features_dict = {}
+    TracksDict = {}
+    for path, im, im0s, vid_cap, s in dataset:
+        if save_img and 'imgshow' not in locals().keys():
+            imgshow = im0s.copy() 
+        
+        ## ============================= tracking 功能只处理视频，writed by WQG
+        if dataset.mode == 'image':
+            continue
+
+        with dt[0]:
+            im = torch.from_numpy(im).to(model.device)
+            im = im.half() if model.fp16 else im.float()  # uint8 to fp16/32
+            im /= 255  # 0 - 255 to 0.0 - 1.0
+            if len(im.shape) == 3:
+                im = im[None]  # expand for batch dim
+
+        # Inference
+        with dt[1]:
+            visualize = increment_path(save_dir / Path(path).stem, mkdir=True) if visualize else False
+            pred = model(im, augment=augment, visualize=visualize)
+
+        # NMS
+        with dt[2]:
+            pred = non_max_suppression(pred, conf_thres, iou_thres, classes, agnostic_nms, max_det=max_det)
+
+        # Process predictions
+        for i, det in enumerate(pred):  # per image
+            seen += 1
+            if webcam:  # batch_size >= 1
+                p, im0, frame = path[i], im0s[i].copy(), dataset.count
+                s += f'{i}: '
+            else:
+                p, im0, frame = path, im0s.copy(), getattr(dataset, 'frame', 0)
+            
+            images.append((dataset.frame, im0))
+            
+            p = Path(p)  # to Path
+            save_path = str(save_dir / p.name)  # im.jpg
+            s += '%gx%g ' % im.shape[2:]  # print string
+            
+            # im0_ant = im0.copy()
+            annotator = Annotator(im0.copy(), line_width=line_thickness, example=str(names))
+            
+            nd = len(det)
+            if nd: 
+                # Rescale boxes from img_size to im0 size
+                det[:, :4] = scale_boxes(im.shape[2:], det[:, :4], im0.shape).round()
+                
+                det = det.cpu().numpy()
+                det = np.concatenate([det[:, :4], np.arange(nd).reshape(-1, 1), det[:, 4:]], axis=-1) 
+
+                DetBoxes = np.concatenate([DetBoxes, det[:, :6]], axis=0)
+                
+                ## ============================================================ 前后帧相同 boxes 的特征赋值 
+                # def static_estimate(box1, box2, TH1=8, TH2=12):
+                #     dij_abs = max(np.abs(box1 - box2))
+                #     dij_euc = max([np.linalg.norm((box1[:2] - box2[:2])), 
+                #                     np.linalg.norm((box1[2:4] - box2[2:4]))
+                #                     ])
+                #     if dij_abs < TH1 and dij_euc < TH2:
+                #         return True
+                #     else:
+                #         return False
+
+                # nw = 3                      # 向前递推检查的窗口大小
+                # nf = len(BoxesFeats)        # 已经检测+特征提取的帧数
+                # feat_curr = [None] * nd     # nd: 当前帧检测出的boxes数
+                # for ii in range(nd):
+                #     box = det[ii, :4]
+                    
+                #     kk=1
+                #     feat = None
+                #     while kk <= nw and nf>=kk:
+                #         ki = -1 * kk
+                #         boxes_ = BoxesFeats[ki][0]
+                #         feats_ = BoxesFeats[ki][1]
+
+                #         flag = [jj for jj in range(len(boxes_)) if static_estimate(box, boxes_[jj, :4])]
+                #         if len(flag) == 1:
+                #             feat = feats_[flag[0]]
+                #             break 
+                #         kk += 1
+                #     if feat is not None:
+                #         feat_curr[ii] = feat
+                        
+                        
+
+                
+                ## ================================================================ writed by WQG
+                
+                '''tracks: [x1, y1, x2, y2, track_id, score, cls, frame_index, box_index]
+                            0   1   2   3      4        5     6       7           8
+                          这里，frame_index 也可以用视频的 帧ID 代替, box_index 保持不变
+                '''
+
+                det_tracking = Boxes(det, im0.shape).cpu().numpy()
+                tracks = tracker.update(det_tracking, im0)
+                if len(tracks) == 0:
+                    continue
+                tracks[:, 7] = dataset.frame
+                
+                '''================== 1. 存储 dets/subimgs/features Dict ============='''
+                imgs, features = inference_image(im0, tracks)
+                
+
+                TrackerFeats = np.concatenate([TrackerFeats, features], axis=0)
+                
+                
+                
+                
+                
+                imgdict = {}
+                boxdict = {}
+                featdict = {}
+                for ii, bid in enumerate(tracks[:, 8]):
+                    imgdict.update({int(bid): imgs[ii]})                    # [f"img_{int(bid)}"] = imgs[i]
+                    boxdict.update({int(bid): tracks[ii, :]})               # [f"box_{int(bid)}"] = tracks[i, :]
+                    featdict.update({int(bid): features[ii, :]})            # [f"feat_{int(bid)}"] = features[i, :]
+                TracksDict[f"frame_{int(dataset.frame)}"] = {"imgs":imgdict, "boxes":boxdict, "feats":featdict}
+                
+                track_boxes = np.concatenate([track_boxes, tracks], axis=0)
+                    
+            
+                '''================== 2. 提取手势位置 ==================='''
+                # idx_0 = tracks[:, 6].astype(np.int_) == 0
+                # hn = 0
+                # for j, index in enumerate(idx_0):
+                #     if index:        
+                #         track = tracks[j, :]
+                #         hand_local, imgshow = handpose.get_hand_local(track, im0)
+                #         handlocals_dict.update({int(track[7]): {int(track[8]): hand_local}})
+                        
+                #         # '''yoloV5和手势检测的召回率并不一直，用hand_local代替tracks中手部的(x1, y1, x2, y2)，会使得两种坐标方式混淆'''
+                #         # if hand_local: tracks[j, :4] = hand_local
+                        
+                #         hn += 1 
+                #         cv2.imwrite(f"D:\DeepLearning\yolov5\hands\images\{Path(source).stem}_{int(track[7])}_{hn}.png", imgshow)
+
+                for *xyxy, id, conf, cls, fid, bid in reversed(tracks):
+                    name = ('' if id==-1 else f'id:{int(id)} ') + names[int(cls)]                   
+                    label = None if hide_labels else (name if hide_conf else f'{name} {conf:.2f}')
+                    
+                    if id >=0 and cls==0:
+                        color = colors(int(cls), True)
+                    elif id >=0 and cls!=0:
+                        color = colors(int(id), True)
+                    else:
+                        color = colors(19, True)     # 19为调色板的最后一个元素
+           
+                    annotator.box_label(xyxy, label, color=color)
+
+            # Save results (image and video with tracking)
+            im0 = annotator.result()
+            save_path_img, ext = os.path.splitext(save_path)
+            if save_img:
+                if dataset.mode == 'image':
+                    imgpath = save_path_img + f"_{dataset}.png"
+                else:
+                    imgpath = save_path_img + f"_{dataset.frame}.png"
+                    
+                cv2.imwrite(Path(imgpath), im0)
+                
+                if vid_path[i] != save_path:  # new video
+                    vid_path[i] = save_path
+                    if isinstance(vid_writer[i], cv2.VideoWriter):
+                        vid_writer[i].release()  # release previous video writer
+                    if vid_cap:  # video
+                        fps = vid_cap.get(cv2.CAP_PROP_FPS)
+                        w = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH))
+                        h = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
+                    else:  # stream
+                        fps, w, h = 30, im0.shape[1], im0.shape[0]
+                    save_path = str(Path(save_path).with_suffix('.mp4'))  # force *.mp4 suffix on results videos
+                    vid_writer[i] = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*'mp4v'), fps, (w, h))
+                vid_writer[i].write(im0)
+
+        # Print time (inference-only)
+        LOGGER.info(f"{s}{'' if len(det) else '(no detections), '}{dt[1].dt * 1E3:.1f}ms")
+
+    
+    ## ======================================================================== written by WQG
+    ## track_boxes: Array, [x1, y1, x2, y2, track_id, score, cls, frame_index, box_id]
+        
+    '''上面保存了检测结果是视频和图像，以下还保存五种类型的数据'''
+    filename = os.path.split(save_path_img)[-1]
+    file, ext = os.path.splitext(filename)
+    
+    mfid = np.max(track_boxes[:, 7])
+    start, end = detect_start_end(track_boxes, features_dict, filename)
+    
+    
+    if start == end:
+        return
+    
+    if start > 5:
+        start = start - 5
+    else:
+        start = 0
+    
+    if mfid - end > 5:
+        end = end + 5
+        
+        
+    # img_path = Path(f'./runs/images/{file}/')
+    
+    img_path = savepath / f'{file}'
+    
+    if not img_path.exists():
+        img_path.mkdir(parents=True, exist_ok=True)
+    
+    '''抽帧间隔数'''
+    Interval = 3
+    for i, img in images:
+        if i >= start and i % Interval == 0:
+            imgpath = img_path / f'{file}_{int(i)}.png'
+            cv2.imwrite(Path(imgpath), img)
+        if i == end:
+            break
+
+    
+    '''======================== 3. save hand_local data =================='''
+    # handlocal_dir = Path('./tracking/data/handlocal/')
+    # if not handlocal_dir.exists():
+    #     handlocal_dir.mkdir(parents=True, exist_ok=True)
+    # handlocal_path = handlocal_dir.joinpath(f'{filename}.pkl')
+    # with open(handlocal_path, 'wb') as file:
+    #     pickle.dump(handlocals_dict, file)
+        
+
+    # Print results
+    t = tuple(x.t / seen * 1E3 for x in dt)  # speeds per image
+    LOGGER.info(f'Speed: %.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape {(1, 3, *imgsz)}' % t)
+    if save_txt or save_img:
+        s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else ''
+        LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}{s}")
+    if update:
+        strip_optimizer(weights[0])  # update model (to fix SourceChangeWarning)
+
+
+def parse_opt():
+    modelpath = ROOT / 'ckpts/best_yolov5m_250000.pt'       # 'ckpts/best_15000_0908.pt',  'ckpts/yolov5s.pt', 'ckpts/best_20000_cls30.pt'
+    
+    '''datapath为视频文件目录或视频文件'''
+    datapath = r"D:/datasets/ym/videos/标记视频/"   # ROOT/'data/videos',  ROOT/'data/images'   images    
+    # datapath = r"D:\datasets\ym\highvalue\videos"
+    # datapath = r"D:/dcheng/videos/"
+    # modelpath = ROOT / 'ckpts/yolov5s.pt' 
+    
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', nargs='+', type=str, default=modelpath, help='model path or triton URL')   # 'yolov5s.pt', best_15000_0908.pt
+    parser.add_argument('--source', type=str, default=datapath, help='file/dir/URL/glob/screen/0(webcam)')    # images, videos
+    parser.add_argument('--data', type=str, default=ROOT / 'data/coco128.yaml', help='(optional) dataset.yaml path')
+    parser.add_argument('--imgsz', '--img', '--img-size', nargs='+', type=int, default=[640], help='inference size h,w')
+    parser.add_argument('--conf-thres', type=float, default=0.25, help='confidence threshold')
+    parser.add_argument('--iou-thres', type=float, default=0.45, help='NMS IoU threshold')
+    parser.add_argument('--max-det', type=int, default=1000, help='maximum detections per image')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--view-img', action='store_true', help='show results')
+    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
+    parser.add_argument('--save-csv', action='store_true', help='save results in CSV format')
+    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')
+    parser.add_argument('--save-crop', action='store_true', help='save cropped prediction boxes')
+    parser.add_argument('--nosave', action='store_true', help='do not save images/videos')
+    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --classes 0, or --classes 0 2 3')
+    parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
+    parser.add_argument('--augment', action='store_true', help='augmented inference')
+    parser.add_argument('--visualize', action='store_true', help='visualize features')
+    parser.add_argument('--update', action='store_true', help='update all models')
+    parser.add_argument('--project', default=ROOT / 'runs/detect', help='save results to project/name')
+    parser.add_argument('--name', default='exp', help='save results to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    parser.add_argument('--line-thickness', default=3, type=int, help='bounding box thickness (pixels)')
+    parser.add_argument('--hide-labels', default=False, action='store_true', help='hide labels')
+    parser.add_argument('--hide-conf', default=False, action='store_true', help='hide confidences')
+    parser.add_argument('--half', action='store_true', help='use FP16 half-precision inference')
+    parser.add_argument('--dnn', action='store_true', help='use OpenCV DNN for ONNX inference')
+    parser.add_argument('--vid-stride', type=int, default=1, help='video frame-rate stride')
+    opt = parser.parse_args()
+    opt.imgsz *= 2 if len(opt.imgsz) == 1 else 1  # expand
+    print_args(vars(opt))
+    return opt
+        
+def find_files_in_nested_dirs(root_dir):
+    all_files = []
+    extensions = ['.mp4']
+    for dirpath, dirnames, filenames in os.walk(root_dir):
+        for filename in filenames:
+            file, ext = os.path.splitext(filename)
+            if ext in extensions:
+                all_files.append(os.path.join(dirpath, filename))
+    return all_files
+            
+print('=======')
+
+def main(opt):
+    check_requirements(ROOT / 'requirements.txt', exclude=('tensorboard', 'thop'))
+       
+    p = r"D:\datasets\ym\永辉测试数据_202404\20240402"
+    
+    optdict = vars(opt)
+    files = []
+    k = 0
+    if os.path.isdir(p):
+        files.extend(sorted(glob.glob(os.path.join(p, '*.*'))))   
+        for file in files:
+            optdict["source"] = file
+            run(**optdict)
+            
+            k += 1
+            if k == 2:
+                break         
+    elif os.path.isfile(p):
+        run(**vars(opt))
+        
+def main_loop(opt):
+    check_requirements(ROOT / 'requirements.txt', exclude=('tensorboard', 'thop'))
+    
+    optdict = vars(opt)   
+    
+    # p = r"D:\datasets\ym\永辉测试数据_比对" 
+    p = r"D:\datasets\ym\广告板遮挡测试\8"
+    # p = r"D:\datasets\ym\videos\标记视频"
+    # p = r"D:\datasets\ym\实验室测试"
+    # p = r"D:\datasets\ym\永辉双摄视频\新建文件夹"
+    
+    k = 0
+    if os.path.isdir(p):
+        files = find_files_in_nested_dirs(p)
+        
+        # files = [r"D:\datasets\ym\广告板遮挡测试\8\6926636301004_20240508-175300_back_addGood_70f754088050_215_17327712807.mp4",
+        #           r"D:\datasets\ym\videos\标记视频\test_20240402-173935_6920152400975_back_174037372.mp4",
+        #           r"D:\datasets\ym\videos\标记视频\test_20240402-173935_6920152400975_front_174037379.mp4",
+        #           r"D:\datasets\ym\广告板遮挡测试\8\2500441577966_20240508-175946_front_addGood_70f75407b7ae_155_17788571404.mp4"
+        #           ]
+        
+        files = [r"D:\datasets\ym\广告板遮挡测试\8\6907149227609_20240508-174733_back_returnGood_70f754088050_425_17327712807.mp4"]
+        
+        
+        for file in files:
+            optdict["source"] = file
+            run(**optdict)
+            
+            k += 1
+            if k == 1:
+                break
+    elif os.path.isfile(p):
+        optdict["source"] = p
+        run(**vars(opt))
+
+
+
+
+if __name__ == '__main__':
+    opt = parse_opt()
+    # main(opt)
+    savepath = ROOT / 'runs/images'
+    main_loop(opt)
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+    
+