DPTracker
March 17, 2026 ยท View on GitHub
A simple and effective tracking framework for nighttime UAV Tracking.
๐ฃ News
- [2026.03] Model checkpoints and raw tracking results are now available!
- [2026.02] Our paper has been accepted by ICRA 2026! ๐
๐ Download
- Model Checkpoints:
- Baidu Netdisk (Code: ac7z)
- Google Drive
- Raw Tracking Results:
- Baidu Netdisk (Code: haax)
- Google Drive
๐ฅ๏ธ Environment
This repo is using Pytorch 2.1.2 with CUDA 11.8.
๐ ๏ธ Setup
Create the default local file with paths:
python tracking/create_default_local_file.py \
--workspace_dir ./ \
--data_dir ./data \
--save_dir ./output
๐ Train
Preparation
Download the training data, including GOT-10K, LASOT, COCO, and TrackingNet, ExDark, Shift, and BDD100K.
Run
conda activate your_env
bash train.sh
๐ Test
Preparation
Download the test data, including NAT2021, UAVDark135, DarkTrack2021.
Run
conda activate your_env
bash test.sh