AlignVTOFF: Texture-Spatial Feature Alignment for High-Fidelity Virtual Try-Off

2026-01-05

arXiv.org

AI Summary

This work addresses the structural distortions and loss of high-frequency texture commonly observed in existing virtual try-on (VTO) methods when handling complex geometric deformations. To mitigate these issues, the authors propose AlignVTO, a novel framework featuring a parallel U-Net architecture. It employs a trainable reference U-Net for multi-scale feature extraction and introduces a Texture-Spatial Feature Alignment (TSFA) mechanism that explicitly aligns garment texture and spatial information. Within a frozen denoising U-Net, TSFA integrates cross-attention and self-attention to achieve precise alignment. Extensive experiments demonstrate that AlignVTO outperforms current state-of-the-art methods across multiple benchmarks, significantly improving both structural accuracy and fidelity of high-frequency details in the generated images.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Multimodal Learning
• Humans and AI: Game Design — Virtual Humans, NPCs and Autonomous Characters

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Virtualization and resource management in Web systems and infrastructures
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Virtual Try-Off (VTOFF) is a challenging multimodal image generation task that aims to synthesize high-fidelity flat-lay garments under complex geometric deformation and rich high-frequency textures. Existing methods often rely on lightweight modules for fast feature extraction, which struggles to preserve structured patterns and fine-grained details, leading to texture attenuation during generation.To address these issues, we propose AlignVTOFF, a novel parallel U-Net framework built upon a Reference U-Net and Texture-Spatial Feature Alignment (TSFA). The Reference U-Net performs multi-scale feature extraction and enhances geometric fidelity, enabling robust modeling of deformation while retaining complex structured patterns. TSFA then injects the reference garment features into a frozen denoising U-Net via a hybrid attention design, consisting of a trainable cross-attention module and a frozen self-attention module. This design explicitly aligns texture and spatial cues and alleviates the loss of high-frequency information during the denoising process.Extensive experiments across multiple settings demonstrate that AlignVTOFF consistently outperforms state-of-the-art methods, producing flat-lay garment results with improved structural realism and high-frequency detail fidelity.

Problem

• Virtual Try-On
• Texture Preservation
• Geometric Deformation
• High-Fidelity Image Generation
• Multimodal Image Synthesis

Innovation

• Texture-Spatial Feature Alignment
• Parallel U-Net
• Hybrid Attention
• High-Fidelity Virtual Try-On
• Reference U-Net