TEST RECORD · ← all field reports

FIELD REPORTS/IMAGE/ISSUE #86

Remove the background from any photo in seconds

Every design tool has a background remover now, and most of them fail the same way: hair turns to mush, bicycle spokes vanish, and anything wispy gets a halo. A small lab called Feyn went after exactly those edges, trained a model on the problem, and published the whole thing: weights, training library, and a demo you can use in the browser.

MODELFeyNobg
PUBLISHEDAugust 10, 2026
READ TIME3 min
TESTED BYNeural Expedition
CATEGORYIMAGE

Field notes

01What it does

You give it an image and it predicts, for every pixel, how much of that pixel belongs to the subject. Background pixels turn transparent, the subject stays solid, and boundary pixels like hair, fur, thin wires, and motion blur become translucent instead of being cut with scissors. That last part is the hard bit, and it is where this model earns its keep: windblown hair and bicycle wheels survive the cutout.

Under the hood it builds on BiRefNet, a well-regarded segmentation architecture, retrained by Feyn on a mix of data designed to balance two skills that usually fight each other: recognizing which pixels form the subject and tracing the boundary precisely. At about 260 million parameters it is compact by modern standards, and the weights are Apache licensed. By the team's own evaluation it posts the best published S-measure on four of eight standard benchmarks, including ultra high resolution images, and stays within two percent of the leader on the rest.

02How to try it

The demo Space is the fastest test: upload a photo that usually breaks removers, someone with flyaway hair, a bike, a chain-link fence, and look at the edges up close. For real use, pip install nobg and about ten lines of Python download the model from Hugging Face and save a cutout PNG. The same NoBg library can also train a background remover from scratch on your own data.

03Caveat

The benchmark wins come from the team's own evaluation, and the model is young enough that independent hands-on reports are still scarce. Its weakest scenes in those same numbers are camouflage-style images where the subject blends into the background, where it trails the best specialist models slightly. For clean product and portrait shots, the common case, it is at or near the top.

04What you can do with it

  • Cut product shots for a store page without reshooting on white.
  • Turn phone photos into sticker packs or profile cutouts with intact hair.
  • Batch clean a whole catalog or thumbnail folder with a short script.
  • Pull subjects for posters and collages without halo artifacts.
  • Retrain the model on your own niche imagery with the open NoBg library.

Try the demo

View model page

Read this issue on neuralexpedition.com →