Detecting pollinators with AI isn't as easy as people say ππ
Studies show high scores on one dataset π then imply the model is useful elsewhere π€₯
@veralis.bsky.social & co. don't board the AI hype train π€π
They expose real challenges when deploying GBIF-trained AI for pollinator surveillance ππͺ°π
Jamie Alison
Our latest research explores how YOLO object detectors, trained on citizen science images, perform on unseen time-lapse images of pollinators captured with a fixed smartphone setup. While successful for larger pollinators, detecting smaller or blurrier flower visitors remains a challenge.
