This repository provides a end-to-end solution for Multi-Object Tracking.
To set up the environment, clone the repository and install the necessary packages:
conda create --name obj_track_conda
conda activate obj_track_conda
pip3 install ipykernel
python3 -m ipykernel install --user --name obj_track_conda --display-name obj_track_conda
pip3 install -r requirements.txt
The demo.mp4 from Internet (Only for demo purpose)
Please prepare your own video clip for demo.
To run the model on a video, use the following command:
python demo.py \
--model_file ./model/object_tracking.onnx \
--video_path ./data/veh2.mp4 \
--line_points 6 793 1675 777 \
--classes_interest person motor-vehicle non-motor-vehicle \
--opts_dir ./res \
--conf_thres 0.1 \
--iou_thres 0.7 \
--draw_tracks \
--draw_boxes
The scripts will save the visualized results.
Note: This is a demonstration. For higher accuracy, please customize the training strategy.
The torch weight is from Yolov8