Arpit Jadon, Valentinos Pariza, Jona Ruthardt, Ryousuke Yamada, A. Sophia Koepke, Andrei Bursuc, Yuki M. Asano (Fundamental AI Lab at UTN) · Open Generative Perception

Open Vision Banana: A Fully Open Recipe for Generative Visual Perception

Open Vision Banana (OVB) converts pretrained image generators into multi-task perception systems by framing vision tasks as conditional image editing, using open 4B and 9B base models and 2.1M-3.8M public training pairs. It reports 73.9 mIoU on Cityscapes zero-shot against Vision Banana's 69.9, and a test-time legend ensembling method that improves mIoU by up to 8.7 points on Cityscapes and 15.0 on COCO-Object without retraining. The authors release the dataset pipeline, training and evaluation code, and final, intermediate and ablation checkpoints; the licence is not stated on the page.

Open the source ↗

Similar Vision & Image releases

  • Rembg

    Daniel Gatis (danielgatis) · Background removal · Week 46, 2026

  • InternLumina-U2

    Shanghai AI Laboratory · Omni-Visual Diffusion LLM · Week 36, 2026

  • ReDesign

    jintae-00 (GitHub) · Visual Understanding · Week 31, 2026

  • MODUS

    EPFL / Apple / U. Copenhagen / CUHK / U. Geneva / Lambda · Visual Understanding · Week 31, 2026

  • Seedream 5.5

    ByteDance Seed · Image generation and editing · Week 47, 2026

  • MiniMax-H3-Image

    MiniMax (MiniMaxAI) · Open 16MP Image Model · Week 47, 2026

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.