Dichotomous image segmentation using bilateral reference modeling
BiRefNet Dis focuses on binary object segmentation tasks where clear separation between subject and background is required. The model emphasizes structural consistency and minimizes foreground leakage through dual reference guidance.
