Postdoc @cbehav.bsky.social, Konstanz 🐦 Trying to track birds with computer vision 🇭🇰
https://alexhhchan.odoo.com/
Alex Chan
Loading...
Traditional computer vision datasets often focus on individual computer vision tasks, but for deployments in biological systems, multiple tasks often need to be combined (e.g., object detection + re-ID + action recognition).
Thank you to all co-authors, and see you in Denver! :)
Co authors: Neha Singhal, Onur Kocahan, Andrea Meltzer, Saverio Lubrano, Miya Warrington, @michael-griesser.bsky.social, Fumihiro Kano, Hemal Naik
Importantly, we provide an "application-specific benchmark", with the goal of benchmarking novel algorithms directly in the context of the final use case (e.g., individual feeding rate + paired co-occurrence rate in this case). We discussed this concept here: arxiv.org/abs/2505.02825
New preprint alert!! Accepted to CVPR 2026, we present the CHIRP dataset, a task-diverse computer vision dataset on Siberian jays. arxiv.org/abs/2603.25524
@cbehav.bsky.social
