This repository contains multiple scripts for various tasks. Each script preforms a separate task. The detections are generated by YOLOv5, a family of object detection architectures and models pretrained on the COCO dataset, are passed to a set of algorithms that track/count/monitor the detected objects. The scripts can perform these operations with any set of yolov5 pretrained weights.
I'll update this repository with more scripts when I am able.
- Clone the repository recursively:
git clone --recurse-submodules https://github.com/Pawan-Valluri/Object-tracing-with-YOLOv5.git
If you already cloned and forgot to use --recurse-submodules you can run git submodule update --init
- Make sure that you fulfill all the requirements: Python 3.8 or later with all requirements.txt dependencies installed, including torch==1.9.0 (with cuda 10.2 or 11.3 for better peformance). To install, run:
pip install -r requirements.txt
These scripts can be run on most video formats
$ python path_counter_per_area_per_class.py --source 0 # webcam
img.jpg # image
vid.mp4 # video
path/ # directory
path/*.jpg # glob
'https://youtu.be/Zgi9g1ksQHc' # YouTube
'rtsp://example.com/media.mp4' # RTSP, RTMP, HTTP streamThese scripts can work an any yolov5 pretrained weights. You can use your weights by executing the following commands.
$ python path_tracer.py --source 0 --yolo_model yolov5n.pt --img 640
yolov5s.pt
yolov5m.pt
yolov5l.pt
yolov5x.pt --img 1280
customyolo.pt
...