Show HN: RelateAnything – Real-time open-vocabulary relation prediction

2 pointsposted 4 hours ago
by maelicn

1 Comments

maelicn

4 hours 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 :)