Vision Banana: How Google DeepMind's Image Generator Beat SAM Three and Depth Anything at Their Own Game - May 1, 2026

Send us Fan Mail Vision Banana: How Google DeepMind's Image Generator Beat SAM Three and Depth Anything at Their Own Game - May 1, 2026 Google DeepMind just published Vision Banana, an instruction tuned image generator built on top of Nano Banana Pro that beats SAM Three on segmentation and Depth Anything Version Three on metric depth. The paper, co-authored by He Kaiming and Xie Saining, argues that image generation pretraining plays the same role for vision that text generation pretraining...