Hasty Briefsbeta

Bilingual

Vision Banana Image Generators Are Generalist Vision Learners

12 hours ago
  • #zero-shot learning
  • #computer vision
  • #image generation
  • Hover over images to reveal Vision Banana's generation results, with mobile requiring tap to toggle.
  • Displays segmentation, instance, referred object masks, depth maps, and surface normal maps on hover.
  • Vision Banana achieves state-of-the-art performance in zero-shot transfer for 2D and 3D vision tasks.
  • Acknowledgments list collaborators and contributors for discussions and guidance.
  • References the article 'Image Generators are Generalist Vision Learners' published in arXiv in 2026.