IGR: Improving Diffusion Model for Garment Restoration from Person Image

TL;DR

This paper introduces an improved diffusion-based method for garment restoration from person images, effectively preserving garment identity and details through novel feature extraction, fusion, and training strategies.

Contribution

It proposes a new diffusion model with dual garment extractors, garment fusion blocks, and a coarse-to-fine training approach for more accurate garment restoration.

Findings

Effective preservation of garment identity in restorations

High-quality results even with complex or occluded garments

Outperforms existing methods in fidelity and authenticity

Abstract

Garment restoration, the inverse of virtual try-on task, focuses on restoring standard garment from a person image, requiring accurate capture of garment details. However, existing methods often fail to preserve the identity of the garment or rely on complex processes. To address these limitations, we propose an improved diffusion model for restoring authentic garments. Our approach employs two garment extractors to independently capture low-level features and high-level semantics from the person image. Leveraging a pretrained latent diffusion model, these features are integrated into the denoising process through garment fusion blocks, which combine self-attention and cross-attention layers to align the restored garment with the person image. Furthermore, a coarse-to-fine training strategy is introduced to enhance the fidelity and authenticity of the generated garments. Experimental results demonstrate that our model effectively preserves garment identity and generates high-quality restorations, even in challenging scenarios such as complex garments or those with occlusions.