Reproduce “UAV-OVVIS: Unmanned Aerial Vehicles Also Need Open-Vocabulary Video Instance Segmentation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mingyu Dou, Shi Qiu, Ming Hu, Yifan Chen, Zhe Sun
AeroTrack combines text queries with pixel-level masks and video-wide identities for UAV video perception. It uses open-vocabulary detections to prompt a segmenter that propagates masks within short intervals, then uses LIA to associate local IDs across restarts. On AeroVIS, YOLO-World + SAM3 scores 55.82 Overall HOTA, versus 34.98 for DEVA; an ablation reports 23.29 without LIA, showing its contribution in this setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.