hjLabs AI Playground
Starting the on-device runtime…
SegFormer-B0 fine-tuned on ADE20K assigns a class to every pixel across 150 scene categories — sky, road, building, person, furniture — and renders the result as a colour-coded overlay with a legend you can toggle per class. Panoptic and instance variants are available where you need separate objects rather than regions. Useful for scene understanding, masking and dataset inspection, all without leaving the browser.
image segmentation onlinesegformer demosemantic segmentation browserade20k modelpixel classification toolpanoptic segmentation onlinefree segmentation ai
Background removal produces one soft alpha matte for a single subject. Segmentation labels every pixel with one of many classes. If you only want a cut-out, use background removal — it is faster and the edges are better.
Yes — each class mask can be downloaded as a PNG, and the full label map is available as raw data.
No. The model weights are downloaded from Hugging Face to your browser once, and every inference after that runs on your own GPU through WebGPU. Your text, images and audio are never sent anywhere — hjLabs.in has no inference server and no way to see your input.
Small models do. Recent mobile Chrome and Safari expose WebGPU, but memory is the binding limit — stay under roughly 200 MB of weights on Android, and expect iOS to reclaim the tab sooner. The catalog shows every download size up front so you can pick something that fits.
That depends on the individual model, not on this site. Each model page links to its Hugging Face repository, where the licence is stated — most of the catalog is Apache-2.0 or MIT, a few are research-only. We add nothing on top: hjLabs.in claims no rights over anything you produce here.