/Deepsort_Yolov5_Pytorch

Combine yolov5 and deepsort to track any project

Primary LanguageJupyter NotebookApache License 2.0Apache-2.0

yolov5 + deepsort

Using yolov5 and deepsort to detect project, which you can choose any class(any class in coco dateset)

Install

  • ubuntu 16.04
  • python >=3.7
  • cuda 10.0
  • cudnn 7.6.5
  • opencv-python 4.2.0.34
  • pytorch 1.5.1
  • torchvision 0.6.1
conda create -n yolov5 python=3.7
conda activate yolov5
git clone https://github.com/Eyren/Deepsort_Yolov5_Pytorch.git Deepsort_yolov5
cd Deepsort_Yolov5_Pytorch-master
pip install -r requirements.txt

Test

  • put the video or image sequence folder you want to test in demo folders.
python yolov5_deepsort.py demo/*.mp4(demo/img) --clsnum 2)

tips: clsnum default is car(2), you can check configs/coco.names to select class name, it count begin 0.(person is 0, bicycle is 1...)

Result

video cut

Contributers

dhpdong, eyren