This work is built on the MCMOT, which adapts to multi-object tracking on moving UAV videos. You can refer to origin fork MCMOT.
i） Prepare dataset
Run gen_dataset_visdrone.py or gen_dataset_UAVDT.py
ii) Run train.py
Run python track.py
The multi-object tracking results saved in **.txt, and evaluate it by the official toolkits.
For visdrone2019 dataset, you can refer to the link(https://github.com/VisDrone/VisDrone2018-MOT-toolkit)
For UAVDT dataset, you can refer to the link(https://github.com/VisDrone/VisDrone2018-MOT-toolkit）
Run python demo.py