Xinlong Li

Let RGB Be the Language of Vision

Timing Yang, Jinrui Yang, Xinlong Li, Yuhan Wang, Haoran Li, Yanqing Liu, et al.

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 visual signals, are all represented as RGB images, while general visual tasks can be converted into a com…Read more