This is a demo for car tracking in a road junction using YOLOv5 (https://docs.ultralytics.com) and view synthesis.

The instructions for creating this repository is available in the Documentation folder.

The demo is availabe on google colab:

Quick Start Examples

Install

Clone repo and install requirements.txt in a Python>=3.7.0 environment, including PyTorch>=1.7.

git clone https://github.com/YJonmo/CarTrackingTask  # clone
cd CarTrackingTask
pip install -r requirements.txt  # install

Inference

Weights:

yolov5n.pt is the fastest but least accurate

yolov5s.pt is slower than 'n' but more accurate

yolov5m.pt is balance between speed and accuracy

yolov5l.pt is more accurate than 'm' but slower

yolov5x.pt is the most accurate model but slowest

Use the following bash command to run the code. You may remove the '--view-img' flag to increase the processing speed.

  
python track_cars.py --yolo-model yolov5s.pt --deep-sort-model osnet_x0_5_market1501 --conf-thres 0.25 --source data/videos --output-path result --view-img