2026
Tuning-Free Amodal Segmentation via the Occlusion-Free Bias of Inpainting Models
AAAI 2026technical
Amodal segmentation is an image-based algorithm that aims to predict masks for both visible and occluded parts of objects. Existing methods typically rely on supervised learning with annotated amodal masks or synthetic data. The effectiveness of these methods relies heavily on the quality of the dat