Forwardrobe: Garment-Aware Gaussian Avatars from a Single Image

2026-07-31

AI Summary

This work addresses the challenge of reconstructing animatable 3D human avatars from a single image under loose clothing—such as skirts or dresses—where existing methods often suffer from inadequate geometric and motion representations. The authors propose a feedforward framework that, for the first time, explicitly models an independently controllable garment layer within a Gaussian avatar formulation, enabling disentangled representations of body and clothing. Key innovations include continuity-aware geometry and skinning initialization, pose-conditioned non-rigid deformations, and appearance-adaptive mechanisms, collectively supporting high-quality animation, garment editing, transfer, and virtual try-on. Experiments demonstrate that the method outperforms current single-image avatar approaches in both reconstruction fidelity and animation quality for loose garments, while significantly enhancing garment manipulability and reusability.

Abstract

Reconstructing animatable 3D human avatars from a single image remains particularly challenging for loose garments, whose geometry and motion cannot be adequately represented by body-aligned topology and skinning. We present Forwardrobe, a feed-forward framework for reconstructing garment-aware Gaussian avatars from a single image. Forwardrobe explicitly separates clothing from the body in canonical Gaussian space and equips the garment layer with continuity-aware geometry and skinning initialization, pose-conditioned non-rigid deformation, and appearance adaptation. These designs improve garment reconstruction and visual quality during animation, particularly for skirts and dresses. The separated garment layer additionally forms an independently controllable 3D asset, enabling garment editing, transfer, and 3D virtual try-on. Experiments demonstrate improved garment reconstruction quality and greater flexibility in garment manipulation compared with existing single-image avatar reconstruction methods.

Problem

• 3D human avatars
• single-image reconstruction
• loose garments
• garment animation
• virtual try-on

Innovation

• garment-aware
• Gaussian avatars
• single-image reconstruction
• non-rigid deformation
• 3D virtual try-on