DrIFT (Autonomous Drone Dataset)

DrIFT: Autonomous Drone Dataset with Integrated Real and Synthetic Data, Flexible Views, and Transformed Domains. This dataset addresses drone detection challenges under domain shifts such as background, point of view, weather, and real-to-synthetic transitions.

Sample & Results Showcase

Paper (arXiv) | Dataset

Samples:

Architecture:

SKY Background:

Tree Background:

Ground Background:

News

• [Apr 2025] The agreement requirement has been released, and you can download it from HuggingFace directly.

• [Dec 2024] The real part of our dataset is in the pipeline and will be online soon.

• [Dec 2024] We have uploaded our dataset V1 (synthetic part) on HuggingFace.

• [Dec 2024] The arXiv version of the paper has been submitted.

• [Oct 2024] Our paper was accepted at IEEE/CVF WACV 2025.

How to download the dataset?

The agreement requirement has been released, and you can download it from HuggingFace directly.

Please download, fill, and sign the agreement and send it to fardad.dadboud@uottawa.ca (please include "DrIFT_download" as the subject of your email). After that, we will pass you permission for the downloading of the dataset from HuggingFace.

Citation

@inproceedings{dadboud2025drift,
  title={DrIFT: Autonomous Drone Dataset with Integrated Real and Synthetic Data, Flexible Views, and Transformed Domains},
  author={Dadboud, Fardad and Azad, Hamid and Mehta, Varun and Bolic, Miodrag and Mantegh, Iraj},
  booktitle={2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
  pages={6900--6910},
  year={2025},
  organization={IEEE}
}