Image Segmentation

Segment images into 150 semantic classes using SegFormer-B0 (15 MB). Identifies objects, landscapes, buildings, and more at the pixel level.

Upload an image to segment

JPG · PNG · WEBP

Powered by Xenova/segformer-b0-finetuned-ade-512-512 via Transformers.js. SegFormer is a lightweight transformer for semantic segmentation with 150 ADE20K classes. Model caches locally after first download. All data stays on your device.