9/ VisionBanana (Google DeepMind)—Unified vision model beats SAM 3 on seg/depth. Prompt-only switching, synthetic-trained. arXiv/demo: https://
buff.ly/kGZUiOC

→ View original post on X — @futurepedia_io