Combined with yolov3, deep sort and flask, it is a target detection and multi-target tracking platform that can run on Web pages. You can upload pictures or videos. When the image is uploaded, target detection is carried out. When the video is uploaded, multi-target tracking is carried out (the default is pedestrian, which can be changed to other objects). The mobile terminal provides an online shooting interface for real-time target detection and multi-target tracking.
- torch
- torchvision
- numpy
- opencv-python==4.1.2.30
- lxml
- tqdm
- flask
- seaborn
- pillow
- vizer
- numba
pip install -r requirements.txt
cd deepsort/detector/YOLOv3/weight/
wget https://pjreddie.com/media/files/yolov3.weights
cd ../../../
cd deep_sort/deep/checkpoint
# download ckpt.t7 from
https://drive.google.com/drive/folders/1xhG0kRH1EX5B9_Iz8gQJb7UNnn_riXi6 to this folder
cd ../../../
yolov3.weight put in deepsort/detector/YOLOv3/weight/
ckpt.t7 put in deep_sort/deep/checkpoint
cd detector/YOLOv3/nms
sh build.sh
cd ../../../
or
cd detector/YOLOv3/nms/ext
python build.py build_ext develop
cd ../../../../
python app.py
7. If you want to configure to run on the server, please visit my blog 阿里云ECS部署python,flask项目,简单易懂,无需nginx和uwsgi
python detector.py
- code: ZQPei/deep_sort_pytorch