BioDrone is the first bionic drone-based single object tracking benchmark, it features videos captured from a flapping-wing UAV system with a major camera shake due to its aerodynamics. BioDrone highlights the tracking of tiny targets with drastic changes between consecutive frames, providing a new robust vision benchmark for SOT.
Main Contributions
Large-scale and high-quality benchmark with robust vision challenges.
We take robust vision research as the entry point to construct BioDrone, which includes 600 videos with 304,209 manually labeled frames, and is annotated and reviewed under a precise process. To our knowledge, BioDrone is the first SOT benchmark collected by the bionic-based vision system and the largest UAV-based SOT benchmark.
Videos from Bionic-based UAV.
Unlike the existing UAV-based datasets that ignore the flapping-wing UAV structure, our team designs the Large Wingspan bionic flight platform with cutting-edge performance for data collection. Compared with other mechanical structures, the flapping-wing system has broader application prospects due to its lifelike bionic structure. Besides, the flapping-wing design includes additional visual challenges due to more damaging camera shake during the air movements.
Rich challenging factor annotation.
Different from existing UAV-based datasets that only provide sequence-level annotation for several challenging factors, BioDrone first provides high-quality fine-grained manual annotations (bounding-box and occlusion annotation) and automatically generate frame-level labels for ten challenge attributes, aiming to provide detailed information for further analyses.
Tracking baselines with comprehensive experimental analyses.
Challenging factors in BioDrone cause algorithms to fail easily. Thus, we optimize the SOTA method KeepTrack and design a new baseline UAV-KT. Besides, we propose a suitable training strategy, and finally achieve a 5% performance boost in the precision score. What's more, BioDrone contains a complete evaluation mechanism and metrics, compares 20 represent methods and 3 proposed baselines, and analyzes their tracking performance in multiple dimensions, aiming to systematically explore the problems of robust vision brought by flapping-wing UAVs.
Latest News
- [2023.10.18] The BioDrone paper has been accepted by International Journal of Computer Vision (IJCV)! The paper, experimental results, and toolkit will be updated gradually.
- [2023.02.16] The biodrone-toolkit has been released!
- [2023.02.15] The submission system and real-time leaderboard have been released!
- [2023.02.14] The download links and the register system have been released!
- [2023.01.10] The home page has been released! More information will be available soon.
Publications
Please cite our paper if BioDrone helps your research.
Demo
Organizers
- Xin Zhao, School of Computer and Communication Engineering, University of Science and Technology Beijing.
- Shiyu Hu, Institute of Automation, Chinese Academy of Sciences.
- Yipei Wang, School of Instrument Science and Engineering, Southeast University.
- Jing Zhang, Institute of Automation, Chinese Academy of Sciences.
- Yimin Hu, Suzhou Institute of Nano-tech and Nano-bionics, Chinese Academy of Sciences.
- Rongshuai Liu, Suzhou Institute of Nano-tech and Nano-bionics, Chinese Academy of Sciences.
- Haibin Ling, Department of Computer Science, Stony Brook University.
- Yin Li, Biostatistics and Medical Informatics Computer Sciences, University of Wisconsin-Madison.
- Renshu Li, Suzhou Institute of Nano-tech and Nano-bionics, Chinese Academy of Sciences.
- Kun Liu, Suzhou Institute of Nano-tech and Nano-bionics, Chinese Academy of Sciences.
- Jiadong Li, Suzhou Institute of Nano-tech and Nano-bionics, Chinese Academy of Sciences.
Maintainer
- Xuchen Li, Institute of Automation, Chinese Academy of Sciences.
Contact
Please contact us if you have any problems or suggestions.