[深度学习]人工智能本科毕业设计基于yolov5的步态识别多目标跨镜头跟踪检测算法系统源码+实现过程

毕业设计题目：基于步态识别的多目标跨镜头跟踪算法研究 主算法：基于yoloV5-deepsort框架进行目标检测和跟踪+GaitSet算法

效果演示：

介绍

此存储库包含一个高度可配置的两阶段跟踪器，可根据不同的部署场景进行调整。YOLOv5是一系列在COCO数据集上预训练的对象检测架构和模型，它生成的检测结果被传递给跟踪对象的深度排序算法。它可以跟踪Yolov5模型训练用来检测的任何对象。

教程

Yolov5自定义数据培训（链接到外部存储库）

DeepSort深度描述符训练（链接到外部存储库）

Yolov5-pytorch评估

在运行跟踪器之前

如果您已经克隆并忘记使用--recurse子模块，则可以运行git-submodule update--init

确保满足所有要求：Python 3.8或更高版本，安装了所有requirements.txt依赖项，包括torch>=1.7。要安装，请运行：

pip install-r requirements.txt

追踪源

视频推理

$ python track.py --source 0  # webcam
                           img.jpg  # image
                           vid.mp4  # video
                           path/  # directory
                           path/*.jpg  # glob
                           'https://youtu.be/Zgi9g1ksQHc'  # YouTube
                           'rtsp://example.com/media.mp4'  # RTSP, RTMP, HTTP stream

选择对象检测和ReID模型

Yolov5

模型推理速度和准确性之间存在明显的权衡。为了满足您的推理速度/准确性需求，您可以选择Yolov5模型进行自动下载

 python track.py --source 0 --yolo_model yolov5n.pt --img 640
                                          yolov5s.pt
                                          yolov5m.pt
                                          yolov5l.pt 
                                          yolov5x.pt --img 1280
                                          ...

DeepSort

以上内容也适用于DeepSort模型。根据您的需求从ReID模型动物园选择ReID模型

python track.py --source 0 --deep_sort_model osnet_x0_5_market1501
                                               resnet50_MSMT17
                                               mobilenetv2_x1_4_dukemtmcreid
                                               ...

筛选跟踪的类

默认情况下，跟踪器跟踪所有MS COCO类。

如果你只想跟踪别人，我建议你使用这些重量来提高表现

python3 track.py --source 0 --yolo_model yolov5/weights/crowdhuman_yolov5m.pt --classes 0  # tracks persons, only

如果要跟踪MS COCO类的子集，请在classes标志后面添加它们相应的索引

python3 track.py --source 0 --yolo_model yolov5s.pt --classes 16 17  # tracks cats and dogs, only

以下是在MS COCO上训练的Yolov5模型可以检测到的所有可能对象的列表。请注意，此repo中类的索引从零开始。

符合MOT的结果

可以通过保存到您的实验文件夹track/exN

python3 track.py--source --save-txt

核心代码track.py

# limit the number of cpus used by high performance libraries
import os
os.environ["OMP_NUM_THREADS"] = "1"
os.environ["OPENBLAS_NUM_THREADS"] = "1"
os.environ["MKL_NUM_THREADS"] = "1"
os.environ["VECLIB_MAXIMUM_THREADS"] = "1"
os.environ["NUMEXPR_NUM_THREADS"] = "1"
import sys
sys.path.insert(0, './yolov5')
import argparse
import os
import platform
import shutil
import time
from pathlib import Path
import cv2
import torch
import torch.backends.cudnn as cudnn
from yolov5.models.experimental import attempt_load
from yolov5.utils.downloads import attempt_download
from yolov5.models.common import DetectMultiBackend
from yolov5.utils.datasets import LoadImages, LoadStreams
from yolov5.utils.general import (LOGGER, check_img_size, non_max_suppression, scale_coords, 
                                  check_imshow, xyxy2xywh, increment_path)
from yolov5.utils.torch_utils import select_device, time_sync
from yolov5.utils.plots import Annotator, colors, save_one_box,save_one_boxmod
from deep_sort.utils.parser import get_config
from deep_sort.deep_sort import DeepSort
FILE = Path(__file__).resolve()
ROOT = FILE.parents[0]  # yolov5 deepsort root directory
if str(ROOT) not in sys.path:
    sys.path.append(str(ROOT))  # add ROOT to PATH
ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative
def detect(opt):
    out, source, yolo_model, deep_sort_model, show_vid, save_vid, save_txt, imgsz, evaluate, half, project, name, exist_ok= \
        opt.output, opt.source, opt.yolo_model, opt.deep_sort_model, opt.show_vid, opt.save_vid, \
        opt.save_txt, opt.imgsz, opt.evaluate, opt.half, opt.project, opt.name, opt.exist_ok
    webcam = source == '0' or source.startswith(
        'rtsp') or source.startswith('http') or source.endswith('.txt')
    device = select_device(opt.device)
    # initialize deepsort
    cfg = get_config()
    cfg.merge_from_file(opt.config_deepsort)
    deepsort = DeepSort(deep_sort_model,
                        device,
                        max_dist=cfg.DEEPSORT.MAX_DIST,
                        max_iou_distance=cfg.DEEPSORT.MAX_IOU_DISTANCE,
                        max_age=cfg.DEEPSORT.MAX_AGE, n_init=cfg.DEEPSORT.N_INIT, nn_budget=cfg.DEEPSORT.NN_BUDGET,
                        )
    # Initialize
    half &= device.type != 'cpu'  # half precision only supported on CUDA
    # The MOT16 evaluation runs multiple inference streams in parallel, each one writing to
    # its own .txt file. Hence, in that case, the output folder is not restored
    if not evaluate:
        if os.path.exists(out):
            pass
            shutil.rmtree(out)  # delete output folder
        os.makedirs(out)  # make new output folder
    # Directories
    if type(yolo_model) is str:
        exp_name = yolo_model.split(".")[0]
    elif type(yolo_model) is list and len(yolo_model) == 1:
        exp_name = yolo_model[0].split(".")[0]
    else:
        exp_name = "ensemble"
    exp_name = exp_name + "_" + deep_sort_model.split('/')[-1].split('.')[0]
    save_dir = increment_path(Path(project) / "out", exist_ok=exist_ok)  # increment run if project name exists
    save_dir.mkdir(parents=True, exist_ok=True)  # make dir
    # Load model
    device = select_device(device)
    model = DetectMultiBackend(yolo_model, device=device, dnn=opt.dnn)
    stride, names, pt, jit, _ = model.stride, model.names, model.pt, model.jit, model.onnx
    imgsz = check_img_size(imgsz, s=stride)  # check image size
    # Half
    half &= pt and device.type != 'cpu'  # half precision only supported by PyTorch on CUDA
    if pt:
        model.model.half() if half else model.model.float()
    # Set Dataloader
    vid_path, vid_writer = None, None
    # Check if environment supports image displays
    if show_vid:
        show_vid = check_imshow()
    # Dataloader
    if webcam:
        show_vid = check_imshow()
        cudnn.benchmark = True  # set True to speed up constant image size inference
        dataset = LoadStreams(source, img_size=imgsz, stride=stride, auto=pt and not jit)
        bs = len(dataset)  # batch_size
    else:
        dataset = LoadImages(source, img_size=imgsz, stride=stride, auto=pt and not jit)
        bs = 1  # batch_size
    vid_path, vid_writer = [None] * bs, [None] * bs
    # Get names and colors
    names = model.module.names if hasattr(model, 'module') else model.names
    # extract what is in between the last '/' and last '.'
    txt_file_name = source.split('/')[-1].split('.')[0]
    txt_path = str(Path(save_dir)) + '/' + txt_file_name + '.txt'
    if pt and device.type != 'cpu':
        model(torch.zeros(1, 3, *imgsz).to(device).type_as(next(model.model.parameters())))  # warmup
    dt, seen = [0.0, 0.0, 0.0, 0.0], 0
    index = 1
    for frame_idx, (path, img, im0s, vid_cap, s) in enumerate(dataset):
        t1 = time_sync()
        img = torch.from_numpy(img).to(device)
        img = img.half() if half else img.float()  # uint8 to fp16/32
        img /= 255.0  # 0 - 255 to 0.0 - 1.0
        if img.ndimension() == 3:
            img = img.unsqueeze(0)
        t2 = time_sync()
        dt[0] += t2 - t1
        # Inference  yolo处理
        visualize = increment_path(save_dir / Path(path[0]).stem, mkdir=True) if opt.visualize else False
        pred = model(img, augment=opt.augment, visualize=visualize)
        t3 = time_sync()
        dt[1] += t3 - t2
        # Apply NMS
        pred = non_max_suppression(pred, opt.conf_thres, opt.iou_thres, opt.classes, opt.agnostic_nms, max_det=opt.max_det)
        dt[2] += time_sync() - t3
        # Process detections
        for i, det in enumerate(pred):  # detections per image
            seen += 1
            if webcam:  # batch_size >= 1
                p, im0, _ = path[i], im0s[i].copy(), dataset.count
                s += f'{i}: '
            else:
                p, im0, _ = path, im0s.copy(), getattr(dataset, 'frame', 0)
            p = Path(p)  # to Path
            save_path = str(save_dir / p.name)  # im.jpg, vid.mp4, ...
            s += '%gx%g ' % img.shape[2:]  # print string
            imc = im0.copy()
            # 画框
            annotator = Annotator(im0, line_width=1, pil=not ascii)
            if det is not None and len(det):
                # Rescale boxes from img_size to im0 size
                det[:, :4] = scale_coords(
                    img.shape[2:], det[:, :4], im0.shape).round()
                # Print results
                for c in det[:, -1].unique():
                    n = (det[:, -1] == c).sum()  # detections per class
                    s += f"{n} {names[int(c)]}{'s' * (n > 1)}, "  # add to string
                xywhs = xyxy2xywh(det[:, 0:4])
                confs = det[:, 4]
                clss = det[:, 5]
                # Deepsort 处理 pass detections to deepsort
                t4 = time_sync()
                outputs = deepsort.update(xywhs.cpu(), confs.cpu(), clss.cpu(), im0)
                t5 = time_sync()
                dt[3] += t5 - t4
                # index = 0
                # for *xyxy, conf, cls in reversed(det):
                #     save_one_box(xyxy, imc, file=save_dir / 'person' /str(output[4])/ f'{index}.jpg', BGR=True)
                #     index = index + 1
                # draw boxes for visualization
                if len(outputs) > 0:
                    for j, (output, conf) in enumerate(zip(outputs, confs)):
                        bboxes = output[0:4]
                        # 这里的ID就是画面中任务头上的ID
                        id = output[4]
                        cls = output[5]
                        c = int(cls)  # integer class
                        label = f'{id} {names[c]} {conf:.2f}'
                        # Todo 保存每一帧检测框中的图片
                        bgrImg = cv2.imread('./data/back.jpg')
                        # print(bgrImg)
                        # save_one_box(bboxes, imc,file=save_dir / 'person' / str(id)/ f'{index}.jpg', BGR=True)
                        save_one_boxmod(bboxes, imc,bgrImg,file=save_dir / 'person' / str(id)/ 'temp'/f'{index}.jpg',
                                        lasted=save_dir / 'person' / str(id)/f'{index}.jpg', BGR=True)
                        index = index+1
                        # 画标记
                        annotator.box_label(bboxes, label, color=colors(c, True))
                        if save_txt:
                            # to MOT format
                            bbox_left = output[0]
                            bbox_top = output[1]
                            bbox_w = output[2] - output[0]
                            bbox_h = output[3] - output[1]
                            # Write MOT compliant results to file
                            with open(txt_path, 'a') as f:
                                f.write(('%g ' * 10 + '\n') % (frame_idx + 1, id, bbox_left,  # MOT format
                                                               bbox_top, bbox_w, bbox_h, -1, -1, -1, -1))
                LOGGER.info(f'{s}Done. YOLO:({t3 - t2:.3f}s), DeepSort:({t5 - t4:.3f}s)')
            else:
                deepsort.increment_ages()
                LOGGER.info('No detections')
            # Stream results
            im0 = annotator.result()
            if show_vid:
                cv2.imshow(str(p), im0)
                if cv2.waitKey(1) == ord('q'):  # q to quit
                    raise StopIteration
            # Save results (image with detections)
            if save_vid:
                if vid_path != save_path:  # new video
                    vid_path = save_path
                    if isinstance(vid_writer, cv2.VideoWriter):
                        vid_writer.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]
                    vid_writer = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*'mp4v'), fps, (w, h))
                vid_writer.write(im0)
    # Print results
    t = tuple(x / seen * 1E3 for x in dt)  # speeds per image
    LOGGER.info(f'Speed: %.1fms pre-process, %.1fms inference, %.1fms NMS, %.1fms deep sort update \
        per image at shape {(1, 3, *imgsz)}' % t)
    if save_txt or save_vid:
        print('Results saved to %s' % save_path)
        if platform == 'darwin':  # MacOS
            os.system('open ' + save_path)
if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--yolo_model', nargs='+', type=str, default='yolov5m.pt', help='model.pt path(s)')
    parser.add_argument('--deep_sort_model', type=str, default='osnet_ibn_x1_0_MSMT17')
    parser.add_argument('--source', type=str, default='0', help='source')  # file/folder, 0 for webcam
    parser.add_argument('--output', type=str, default='inference/output', help='output folder')  # output folder
    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=0.5, help='object confidence threshold')
    parser.add_argument('--iou-thres', type=float, default=0.5, help='IOU threshold for NMS')
    parser.add_argument('--fourcc', type=str, default='mp4v', help='output video codec (verify ffmpeg support)')
    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
    parser.add_argument('--show-vid', action='store_true', help='display tracking video results')
    parser.add_argument('--save-vid', action='store_true', help='save video tracking results')
    parser.add_argument('--save-txt', action='store_true', help='save MOT compliant results to *.txt')
    # class 0 is person, 1 is bycicle, 2 is car... 79 is oven
    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --class 0, or --class 16 17')
    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('--evaluate', action='store_true', help='augmented inference')
    parser.add_argument("--config_deepsort", type=str, default="deep_sort/configs/deep_sort.yaml")
    parser.add_argument("--half", action="store_true", help="use FP16 half-precision inference")
    parser.add_argument('--visualize', action='store_true', help='visualize features')
    parser.add_argument('--max-det', type=int, default=1000, help='maximum detection per image')
    parser.add_argument('--dnn', action='store_true', help='use OpenCV DNN for ONNX inference')
    parser.add_argument('--project', default=ROOT / 'runs/track', 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')
    opt = parser.parse_args()
    opt.imgsz *= 2 if len(opt.imgsz) == 1 else 1  # expand
    with torch.no_grad():
        detect(opt)

完整代码下载地址：https://download.csdn.net/download/FL1768317420/89232934