Skip to content

zhaozijie2022/reID

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 

Repository files navigation

reid_course_project_term7

使用Yolov4+PCB_RPP算法完成行人重识别课程设计

下载数据集及权重

  1. Market-1501-v15.09.15.zip:https://drive.google.com/file/d/0B8-rUzbwVRk0c054eEozWG9COHM/view?resourcekey=0-8nyl7K9_x37HlQm34MmrYQ to ./Dataset
  2. yolov4.weights:https://drive.google.com/open?id=1cewMfusmPjYWbrnuJRuKhPMwRe_b9PaT to ./Yolov4-detector/data/
  3. UESTC-ReID:https://pan.baidu.com/s/1VZPfZOT2Ig6-rZD04Fx5zw (提取码:reid) to ./Dataset/UESTC-ReID

文件介绍

  1. Dataset/UESTC-ReID:

    check_label.py用于提出gt中的坏bbox,并将结果储存在label_check文件夹中
    query_reorganize.py用于将query重新排列为torch.utils.data.Dataloader规定的形式:./id/imgs.jpg
    get_gallery_gt.py根据label从逐帧的img中抠出行人图像,命名规则为:gallery/id/id_camx_xxx.jpg id为两位数(01-17),xxx为第几帧
    sample_gallery.py从每个id每个cam中采样num_img张图片,减小gallery规模以加快test速度

  2. Yolov4-detector:

    detect_Yolo.py 读取数据集的图片, 使用Yolo检测,标注出所有行人并生成txt文件(img_path, res_path) \ label_bbox.py 对 detect_Yolo 得到的 bbox 打标签: 计算与 gt 的 iou 并给出标签 (gt_path, bbox_path, iou_bias, res_path)
    get_gallery_detect.py 根据res_detect给出的label, 在img中抠出图像, 命名规则同上(img_path, label_path, save_path)
    res_disp.py 从detect和gt两种gallery中随机采样相同名称的img展示, 查看detector效果
    models.py from https://github.com/Tianxiaomo/pytorch-YOLOv4

  3. PCB_RPP-classifier

    mylib4test.py test所需的函数库,具体见注释
    get_features.py 对gallery和query提取特征, 并使用npy格式存储
    res_disp.py 随机抽取num_img张query,展示效果
    evaluate_test.py 根据features, 计算CMC & mAP train.py, prepare.py & model.py from https://github.com/Xiaoccer/ReID-PCB_RPP

效果展示

det1
detector效果

hit1
classifier效果

About

行人重识别, 本科第7学期课程设计

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages