Feyn open-sources FeyNoBg, a background-removal model and training library
snyy · hn · 2026-07-28
Feyn is releasing FeyNoBg, an automatic background-removal model, together with NoBg, the Python library used to train and run it.
The team says background removal is deceptively hard: the model must both identify the foreground and trace precise boundaries with per-pixel opacity. They describe controlled evaluations showing that training-data mix can trade off foreground localization against boundary quality—for example, a run trained only on MaskFactory improved on CAMO but regressed on DIS5K.
Their approach is interpretability-first: they studied how BiRefNet’s stages contribute to foreground detection and boundary reconstruction, and found the third stage of the feature extractor carries a large amount of useful information. The post also links to a Hugging Face demo, GitHub library, sample outputs, and a live demo video.
More from Multimodal
- Third-party test: Claude Opus 5.5 renders finer 3D scenes but costs 13x more than GPT-6 Sol — testingcatalog · 2026-09-23
- ComfyUI trick: aux preprocessor + Qwen transfers poses across characters with one prompt — Acceptable-Work8202 · 2026-09-23
- Same portrait prompt across Midjourney V6.1, V7 and V8.2: do older models look better? — tisch_eins · 2026-09-23
- Testing AI character consistency across a 20-image travel sequence — SiennaVaire · 2026-09-23
- Midjourney v8.2 Faces: New Portrait Generation Samples Shared — azed_ai · 2026-09-23
- One-sentence prompt generates lifelike dog video, shown side-by-side with the real one — wgrathwohl · 2026-09-23