Let RGB Be the Language of Vision

Timing Yang, Jinrui Yang, Xinlong Li +11

This work introduces a unified formulation for vision models, where diverse forms of visual information beyond natural images, such as masks, depth maps, and other structured visua…