Evidential Detection and Tracking Collaboration: New Problem, Benchmark and Algorithm for Robust Anti-UAV System

Xue-Feng Zhu, Tianyang Xu, Jian Zhao, Jia-Wei Liu, Kai Wang, Gang Wang, Jianan Li, Qiang Wang, Lei Jin, Zheng Zhu, Junliang Xing, Xiao-Jun Wu*

[Models and Raw results] (Google Driver)

[Paper] (ArXiv)

The dataset will be released soon.

Install the environment

Use the Anaconda

conda create -n edtc python=3.6
conda activate edtc
bash install_pytorch17.sh

Training

Train YOLO

Edit dataset settings:

/path/to/EDTC/yolov5/data/antiuav.yaml

Then train the detector:

cd /path/to/EDTC/yolov5
python train.py

Train tracking branch

Run the following command to set paths for this project

python tracking/create_default_local_file.py --workspace_dir . --data_dir ./data --save_dir .

After running this command, you can modify paths by editing the file:

lib/train/admin/local.py  # paths about training

experiments/uavtrack/baseline.yaml # paths about Stage1 training

experiments/uavtrack_eh/baseline.yaml # paths about Stage2 training

Stage1 training:

python tracking/train.py --script uavtrack --config baseline --save_dir . --mode multiple --nproc_per_node 8

Stage2 training:

python tracking/train.py --script uavtrack_eh --config baseline --save_dir /PATH/TO/SAVE/UAVTRACK_EH --mode multiple --nproc_per_node 8 --stage1_model /STAGE1/MODEL/PATH

Evaluation on AntiUAV600

Download the pretrained models [Models and Raw results] (Google Driver)

Edit the file:

lib/test/evaluation/local.py  # paths about testing
./tracking/test.py

Edit the Line 133-134 of the files, set them according to the paths of YOLO model:

experiments/uavtrack_eh/baseline.yaml  # YOLO pretrained model path

Before run the model:

export PYTHONPATH=$PYTHONPATH:/path/to/EDTC
export PYTHONPATH=$PYTHONPATH:/path/to/EDTC/yolov5

Then run the model:

python tracking/test.py uavtrack_eh baseline --dataset antiuav --threads 32 --num_gpus 8 --params__model /path/to/UAVTrackEH.pth.tar --params__search_area_scale 4.55

If you want to visualize the tracking results, please set:

self.show_result = True   # /path/to/EDTC/lib/test/evaluation/environment.py Line 27 

Evaluation measure

python tracking/evaluate_antiuav_performance.py