ObjectClear: Complete Object Removal via Object-Effect Attention

2025-05-28

AI Summary

To address artifacts, background distortion, and hallucinated content arising from joint removal of target objects and their visual effects (e.g., shadows, reflections) in images, this paper proposes ObjectClear—a diffusion-based end-to-end framework. Its core contributions are threefold: (1) the first object-effect attention mechanism, enabling precise decoupling of foreground removal and background reconstruction; (2) OBER—the first high-quality paired dataset encompassing diverse real and synthetic multi-object complex scenes; and (3) an attention-guided adaptive feature fusion strategy to enhance background fidelity. ObjectClear integrates attention mask learning with paired supervised training. Extensive experiments demonstrate that ObjectClear significantly outperforms state-of-the-art methods on complex scenes: artifact count decreases by 32%, FID improves by 19%, object and effect removal is more complete, and background detail preservation is superior.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Adversarial Learning & Robustness
• Intelligent Robots: Manipulation

Application Category

• Web Mining and Content Analysis: Web data integration and cleaning
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

Object removal requires eliminating not only the target object but also its effects, such as shadows and reflections. However, diffusion-based inpainting methods often produce artifacts, hallucinate content, alter background, and struggle to remove object effects accurately. To address this challenge, we introduce a new dataset for OBject-Effect Removal, named OBER, which provides paired images with and without object effects, along with precise masks for both objects and their associated visual artifacts. The dataset comprises high-quality captured and simulated data, covering diverse object categories and complex multi-object scenes. Building on OBER, we propose a novel framework, ObjectClear, which incorporates an object-effect attention mechanism to guide the model toward the foreground removal regions by learning attention masks, effectively decoupling foreground removal from background reconstruction. Furthermore, the predicted attention map enables an attention-guided fusion strategy during inference, greatly preserving background details. Extensive experiments demonstrate that ObjectClear outperforms existing methods, achieving improved object-effect removal quality and background fidelity, especially in complex scenarios.

Problem

• Removing objects and their effects accurately
• Addressing artifacts in diffusion-based inpainting methods
• Preserving background details during object removal

Innovation

• Object-effect attention mechanism for removal guidance
• Attention-guided fusion preserves background details
• OBER dataset with paired images and precise masks