MOT-tools

Installation

Step1. Install MOT-tools.

git clone https://github.com/guanzhiyu817/MOT-tools.git
cd MOT-tools
pip3 install -r requirements.txt
python3 setup.py develop

Step2. Install pycocotools.

pip3 install cython; pip3 install 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'

Step3. Others

pip3 install cython_bbox

Data preparation

Download MOT17, MOT20 , DanceTrackand put them under /datasets in the following structure:

datasets  
|——————MOT17  
| └——————train  
| └——————test  
└——————MOT20  
| └——————train  
| └——————test
└——————dancetrack        
| └——————train
| └——————val  
| └——————test

Other datasets

Tools

convert_mot17_to_coco.py

Converting MOT17 dataset to coco format

python3 tools/convert_mot17_to_coco.py

convert_mot20_to_coco.py

Converting MOT20 dataset to coco format

python3 tools/convert_mot20_to_coco.py

convert_dance_to_coco.py

Converting DanceTrack dataset to coco format

python3 tools/convert_dance_to_coco.py

convert_video.py

Use the video xxx.mp4 to get xxx_converted.mp4, the new video file has the same size and frame rate as the original video

python3 tools/convert_video.py

interpolation.py

Evaluation of object tracking results and interpolation of trajectories

python tools/interpolation.py

mota.py

Evaluation of tracking results

trt.py

Converting Pytorch Models to TensorRT Models

python tools/trt.py -c best_ckpt.pth.tar

txt2video.py

There are two functions, txt2img and img2video, which convert a text file containing ground-truth information into an image, and the resulting image sequence into a video file, respectively