BiRefNet

Produces one high-resolution foreground mask per image, holding edges on thin structures.

Field
Cross-domain
Method
Deep learning
Data it takes
imagery
License
MIT

BiRefNet returns the salient foreground of an image at full resolution, with clean boundaries on hair, twigs, antennae and grain edges that coarser segmenters smear.

It takes no prompt and no class list, and it returns a single mask. That suits separating a specimen from its background before measurement, or cutting an object out of a field photograph.

Where the target has to be named, Grounding DINO and SAM 2 are the prompted alternatives in the catalog.

Upstream project → · Paper →

Catalog entry last checked 2026-09-08. All models →

Run BiRefNet on your data

The lab runs this on its own compute, in a container, with the inputs and parameters recorded alongside the result. Initial scoping conversations are free.

Contact the Lab