Sequential Amodal Segmentation via Cumulative Occlusion Learning

TL;DR

This paper presents a diffusion-based model for sequential amodal segmentation that effectively predicts complete object contours and occlusion order in complex scenes, surpassing existing methods.

Contribution

The introduction of a cumulative occlusion learning diffusion model enables robust segmentation of occluded objects across diverse categories, advancing beyond prior class-limited approaches.

Findings

Outperforms baseline methods on three amodal datasets.

Effectively captures uncertainty in invisible regions.

Accurately predicts occlusion order and complete object contours.

Abstract

To fully understand the 3D context of a single image, a visual system must be able to segment both the visible and occluded regions of objects, while discerning their occlusion order. Ideally, the system should be able to handle any object and not be restricted to segmenting a limited set of object classes, especially in robotic applications. Addressing this need, we introduce a diffusion model with cumulative occlusion learning designed for sequential amodal segmentation of objects with uncertain categories. This model iteratively refines the prediction using the cumulative mask strategy during diffusion, effectively capturing the uncertainty of invisible regions and adeptly reproducing the complex distribution of shapes and occlusion orders of occluded objects. It is akin to the human capability for amodal perception, i.e., to decipher the spatial ordering among objects and accurately predict complete contours for occluded objects in densely layered visual scenes. Experimental results across three amodal datasets show that our method outperforms established baselines.