Hacker News

maelicn
Show HN: RelateAnything – Real-time open-vocabulary relation prediction github.com

maelicnop7 days ago

The first (and only) model for predicting any relations between any regions (boxes, masks) of an image. It runs localy on your browser or on a GPU under 20ms. This model is my biggest release so far, made entirely on my free time during the past 6 months. I would greatly appreciate your feedback! Please don't hesitate to try the browser demo :)

hn-front (c) 2024 voximity
source