Open-World Amodal Appearance Completion

TL;DR

This paper presents a training-free, open-world amodal appearance completion framework that uses flexible text queries to reconstruct occluded objects in diverse scenes, unifying segmentation, occlusion analysis, and inpainting.

Contribution

It introduces a novel reasoning-based amodal completion method that generalizes to arbitrary objects and queries without training, advancing open-world scene understanding.

Findings

Effective generalization to novel objects and occlusions

Unifies segmentation, occlusion analysis, and inpainting

Establishes a new benchmark for open-world amodal completion

Abstract

Understanding and reconstructing occluded objects is a challenging problem, especially in open-world scenarios where categories and contexts are diverse and unpredictable. Traditional methods, however, are typically restricted to closed sets of object categories, limiting their use in complex, open-world scenes. We introduce Open-World Amodal Appearance Completion, a training-free framework that expands amodal completion capabilities by accepting flexible text queries as input. Our approach generalizes to arbitrary objects specified by both direct terms and abstract queries. We term this capability reasoning amodal completion, where the system reconstructs the full appearance of the queried object based on the provided image and language query. Our framework unifies segmentation, occlusion analysis, and inpainting to handle complex occlusions and generates completed objects as RGBA elements, enabling seamless integration into applications such as 3D reconstruction and image editing. Extensive evaluations demonstrate the effectiveness of our approach in generalizing to novel objects and occlusions, establishing a new benchmark for amodal completion in open-world settings. The code and datasets will be released after paper acceptance.