RelateAnything: 53M-parameter model detects object relations in 20ms

Open-source model RelateAnything, with just 53M parameters, identifies open-vocabulary object relations (holding, wearing, sitting on) from pixels and bounding boxes in about 20ms.

2026-09-19 ~ 2026-09-19 · 2 related posts

• RelateAnything detects object relations from pixels with a 53M-parameter model at ~20ms per frame — Roger_M_Taylor · 2026-09-19
• RelateAnything: a 53M-parameter open-vocabulary CV model that spots object relations in 20ms — blaizedsouza · 2026-09-19