bio-oid/ByteTrack_yolov5

ByteTrack_yolov5

This is an implement of ByteTrack with YOLOv5 backbone instead of YOLOX.

All code is borrowed from ByteTrack and YOLOv5.

Usage

1. Install packages:

   pip install -r requirements.txt

2. Download pretrain model from yolov5 model zoo.
3. Run the code:

   %cd yolov5
   python demo_track.py 
       <demo type, eg. image, video and webcam>
       --path # path to images or video
       --device # gpu or cpu
       --num_classes
       -c,--ckpt # path to checkpoint
       --tsize # input size, only (608, 1088) for yolov5s,m,n and (800, 1440) for yolov5l,x

   example:

   python demo_track.py image --path ./data --device gpu --num_classes 80 --ckpt ./weights/yolov5s.pth --tsize (608, 1088)