Show HN: FeyNoBg – Automatic background removal model and training library

Hacker News is reporting on Feyn's release of FeyNoBg, an open-source model for automatic background removal designed to handle edge cases like fine hair and camouflage. Alongside it, the company is open-sourcing NoBg, a Python library that unifies fragmented image matting tools under one interface. Rather than trading off foreground detection for boundary precision, Feyn took an interpretability-first approach, discovering that one layer of the model's architecture contains most of the information needed for both tasks. Trained on over twenty-six thousand diverse examples, FeyNoBg ranks first on four published benchmarks and comes within two percent on the remaining evaluations.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton