AerialMind: Towards Referring Multi-Object Tracking in UAV Scenarios
Chenglizhao Chen, Shaofeng Liang, Runwei Guan, Xiaolou Sun, Haocheng Zhao, Haiyun Jiang, Tao Huang, Henghui Ding, Qing-Long Han · 2025
Read AerialMind: Towards Referring Multi-Object Tracking in UAV Scenarios by Chenglizhao Chen, Shaofeng Liang, Runwei Guan, Xiaolou Sun, Haocheng Zhao, Haiyun Jiang, Tao Huang, Henghui Ding, Qing-Long Han (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-21053.txt.
Description
Chenglizhao Chen and colleagues introduce in 2025 AerialMind, the first benchmark of its scale for referring multi-object tracking from drones. The task: given a spoken-style instruction, find and follow the matching objects through video. Existing datasets cover ground-level views, but drone footage adds fast motion, camera rotation, tiny objects, and wide scenes. AerialMind contains 48,485 frames with an average of 247.4 language expressions per sequence, second-highest among benchmarks.
Building it by hand would have been prohibitively slow, so the team created COALA, a semi-automated labelling assistant where AI agents draft annotations for human checking. They also propose HawkEyeTrack, a tracking method tuned for aerial perception, and two new metrics isolating scene-induced and motion-induced failures. The benchmark bites: an established transformer tracker scored a HOTA of 19.51 in-domain, where higher means better tracking. Experiments, the authors report, confirm both the dataset's difficulty and HawkEyeTrack's effectiveness.
Catalog record
| CREATOR | Chenglizhao Chen, Shaofeng Liang, Runwei Guan, Xiaolou Sun, Haocheng Zhao, Haiyun Jiang, Tao Huang, Henghui Ding, Qing-Long Han |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 56 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-21053