人流量统计、特定目标搜索+跟踪！

yolov5
deepsort
fastreid

前言

基于YOLOv5和DeepSort的目标跟踪介绍过针对行人的检测与跟踪。本文介绍另一个项目，结合 FastReid 来实现行人的检测、跟踪和重识别。项目地址: https://github.com/zengwb-lx/Yolov5-Deepsort-Fastreid，作者给出的2个主要实例，也是非常的实用，包括行人流量的统计、人群中特定目标的查找与跟踪。

项目复现

首先，创建个全新的虚拟环境

conda create -n pytorch1.6 python=3.7
conda activate pytorch1.6

接着去拉取源码

git clone https://github.com/zengwb-lx/Yolov5-Deepsort-Fastreid.git
cd Yolov5-Deepsort-Fastreid

然后安装下其它的依赖包

# 如果没有gpu的话，就按照requirements.txt安装即可
pip install torch==1.6.0+cu101 torchvision==0.7.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html# 编辑requirements.txt，注释掉torch和torchvision
pip install -r requirements.txt# 使用cython加速
pip install cython
cd fast_reid/fastreid/evaluation/rank_cylib
make all
cd ../../../../

先来跑个行人计数的demo

python person_count.py

yolov5 deepsort fastreid

yolov5的作者分别在googleapi和github上都保存了模型文件，但是这2个yolov5s.pt是不一样的，大家可以通过md5sum去查看一下，github上的模型文件是对的

如果你在运行过程中出现下面的错误

2021-07-13 14:22:20 [INFO]: Loading weights from ./deep_sort/deep/checkpoint/ckpt.t7... Done!
Traceback (most recent call last):File "person_count.py", line 244, in <module>yolo_reid.deep_sort()File "person_count.py", line 121, in deep_sortbbox_xywh, cls_conf, cls_ids, xy = self.person_detect.detect(video_path, img, ori_img, vid_cap)File "/home/xugaoxiang/Works/Yolov5-Deepsort-Fastreid/person_detect_yolov5.py", line 95, in detectpred = self.model(img, augment=self.augment)[0]File "/home/xugaoxiang/anaconda3/envs/pytorch1.6/lib/python3.7/site-packages/torch/nn/modules/module.py", line 722, in _call_implresult = self.forward(*input, **kwargs)File "/home/xugaoxiang/Works/Yolov5-Deepsort-Fastreid/models/yolo.py", line 111, in forwardreturn self.forward_once(x, profile)  # single-scale inference, trainFile "/home/xugaoxiang/Works/Yolov5-Deepsort-Fastreid/models/yolo.py", line 131, in forward_oncex = m(x)  # runFile "/home/xugaoxiang/anaconda3/envs/pytorch1.6/lib/python3.7/site-packages/torch/nn/modules/module.py", line 722, in _call_implresult = self.forward(*input, **kwargs)File "/home/xugaoxiang/Works/Yolov5-Deepsort-Fastreid/models/yolo.py", line 36, in forwardself.training |= self.exportFile "/home/xugaoxiang/anaconda3/envs/pytorch1.6/lib/python3.7/site-packages/torch/nn/modules/module.py", line 772, in __getattr__type(self).__name__, name))
torch.nn.modules.module.ModuleAttributeError: 'Detect' object has no attribute 'export'

这个就是模型的问题，建议使用源码中自带的shell脚本进行下载

sh weights/download_weights.sh

我们来看看这个行人流量统计的基本原理：

首先，作者将yolov5的目标检测封装成了一个类Person_detect，通过它的detect方法可以检测到视频中的每一个行人目标

然后，在画面中设定一条基准线，给定线条两端的坐标即可

line = [(0, int(0.48 * ori_img.shape[0])), (int(ori_img.shape[1]), int(0.48 * ori_img.shape[0]))]
cv2.line(ori_img, line[0], line[1], (0, 255, 255), 4)

接着，创建跟踪器，开始对yolov5检测出的每一个目标进行跟踪。这里以目标预测框的中心点为基准，下图是它的计算方法

yolov5 deepsort fastreid

如果前后帧的中心点所连成的直线和预先设定的基准线相交，则判定为越线，但是这里还有个方向的问题，向上还是向下？来看另一张图

yolov5 deepsort fastreid

作者利用了三角形的正切与反正切原理，使用math模块中的degrees方法来判断，如果这个角度 >0，说明是向上走，反之则为向下走

def vector_angle(midpoint, previous_midpoint):x = midpoint[0] - previous_midpoint[0]y = midpoint[1] - previous_midpoint[1]return math.degrees(math.atan2(y, x))

看完行人计数的示例，我们再来看看特定目标的重识别示例

python person_search_reid.py

报错了

Fusing layers...
Traceback (most recent call last):File "person_search_reid.py", line 120, in <module>yolo_reid = yolo_reid(cfg, args, path=args.video_path)File "person_search_reid.py", line 35, in __init__self.deepsort = build_tracker(cfg, args.sort, use_cuda=use_cuda)File "/home/xugaoxiang/Works/Yolov5-Deepsort-Fastreid/deep_sort/__init__.py", line 18, in build_trackermax_age=cfg.DEEPSORT.MAX_AGE, n_init=cfg.DEEPSORT.N_INIT, nn_budget=cfg.DEEPSORT.NN_BUDGET, use_cuda=use_cuda)File "/home/xugaoxiang/Works/Yolov5-Deepsort-Fastreid/deep_sort/deep_reid.py", line 29, in __init__self.extractor = Reid_feature()File "./fast_reid/demo/demo.py", line 84, in __init__cfg = setup_cfg(args)File "./fast_reid/demo/demo.py", line 35, in setup_cfgcfg.merge_from_file(args.config_file)File "./fast_reid/fastreid/config/config.py", line 107, in merge_from_filecfg_filename, allow_unsafe=allow_unsafeFile "./fast_reid/fastreid/config/config.py", line 50, in load_yaml_with_basewith PathManager.open(filename, "r") as f:File "./fast_reid/fastreid/utils/file_io.py", line 357, in openpath, mode, buffering=buffering, **kwargsFile "./fast_reid/fastreid/utils/file_io.py", line 251, in _openopener=opener,
FileNotFoundError: [Errno 2] No such file or directory: '../../kd-r34-r101_ibn/config-test.yaml'

这是缺少配置文件，到下面的链接去下载

链接: https://pan.baidu.com/s/1bMG3qy7npecCh6AzNO-Zyw
提取码: hy1m

把这2个文件都存放在目录kd-r34-r101_ibn下，然后修改源码fast_reid/demo/demo.py中的第45行，将

default='../../kd-r34-r101_ibn/config-test.yaml',

改成

default='kd-r34-r101_ibn/config-test.yaml',

将第68行的

default=['MODEL.WEIGHTS', '../../kd-r34-r101_ibn/model_final.pth'],

改成

default=['MODEL.WEIGHTS', 'kd-r34-r101_ibn/model_final.pth'],

然后再次运行脚本person_search_reid.py，可以得到

yolov5 deepsort fastreid

可以看到，由于事先已经提了2位行人(a1111111111和b222222222222)的特征，所以，画面中能够识别出这2个人并进行跟踪。默认，特征文件保存在fast_reid/query下

特征提取

如果你也想要制作一个特征文件，可以按照下面的步骤进行

首先，需要截取目标人物的图片，存放在某个以特定目标命名的文件夹下，如我这里的xugaoxiang.com，这样，后面进行识别的时候，就显示xugaoxiang.com这个名字了。把这个文件夹拷贝到fast_reid/query目录下，目录结构如下

(pytorch1.6) xugaoxiang@1070Ti:~/Works/Yolov5-Deepsort-Fastreid/fast_reid/query$ tree
.
├── names.npy
├── query_features.npy
└── xugaoxiang.com├── 10.png├── 11.png├── 12.png├── 13.png├── 14.png├── 15.png├── 1.png├── 2.png├── 3.png├── 4.png├── 5.png├── 6.png├── 7.png├── 8.png└── 9.png

接下来执行

cd fast_reid/demo
python person_bank.py

执行完毕后，query目录下的query_features.npy和names.npy就被更新了

最后，找个包含目标的视频测试下效果

yolov5 deepsort fastreid

工程下载

最后，将包含所需文件全部打包，需要的朋友自行下载

链接：https://pan.baidu.com/s/1JiFzo5_H7TKniZZPzjZK7A
提取码：nauu

参考资料

https://github.com/zengwb-lx/Yolov5-Deepsort-Fastreid
https://blog.csdn.net/zengwubbb/article/details/113422048
https://xugaoxiang.com/2021/06/29/fastreid/

yunanwen