← ArchivePaper2024
SwinGar: Spectrum-Inspired Neural Dynamic Deformation for Free-Swinging Garments
Frequency-domain supervision enables a unified neural model to generate dynamic deformations for garments with arbitrary topology and looseness.
Abstract
Our work presents a novel spectrum-inspired learning-based approach for generating clothing deformations with dynamic effects and personalized details. Existing methods in the field of clothing animation are limited to either static behavior or specific network models for individual garments, which hinders their applicability in real-world scenarios where diverse animated garments are required. Our proposed method overcomes these limitations by providing a unified framework that predicts dynamic behavior for different garments with arbitrary topology and looseness, resulting in versatile and realistic deformations. First, we observe that the problem of bias towards low frequency always hampers supervised learning and leads to overly smooth deformations. To address this issue, we introduce a frequency-control strategy from a spectral perspective that enhances the generation of high-frequency details of the deformation. In addition, to make the network highly generalizable and able to learn various clothing deformations effectively, we propose a spectral descriptor to achieve a generalized description of the global shape information. Building on the above strategies, we develop a dynamic clothing deformation estimator that integrates graph attention mechanisms with long short-term memory.
How to read this
- Category
- Method: learning-based dynamic garment deformation
- Contributions
- A unified neural framework predicting dynamic deformation for garments with arbitrary topology and looseness
- A frequency-control strategy from a spectral perspective to counter low-frequency bias and recover high-frequency detail
- A spectral descriptor giving a generalized global shape description for cross-garment generalization
- Context
- Builds on learning-based clothing animation such as Santesteban et al.'s virtual try-on work, generalizing beyond per-garment networks and static behavior toward one model for diverse free-swinging garments.Builds on: Learning-Based Animation of Clothing for Virtual Try-On
- Correctness
- Rests on the observation that supervised learning is biased toward low frequencies and produces oversmoothed results; the spectral strategy targets that, but as a learned approximation it is validated on the authors' garment set and one should check how it holds for unseen topologies and extreme looseness.
- Clarity
- Moderately technical due to the spectral framing; a first pass conveys the motivation, a second pass is needed for the frequency-control and descriptor formulation.
- How to read it
- Focus on why low-frequency bias matters and how the spectral descriptor enables topology-agnostic generalization; a second pass on the frequency-domain supervision pays off if you work on neural cloth.
Builds on
Built upon by
Nothing yet.
Related work
- PBNS: Physically Based Neural Simulation for Unsupervised Garment Pose Space Deformation 2021 / SIGGRAPH Asia
- SNUG: Self-Supervised Neural Dynamic Garments 2022 / CVPR
- A Pixel-Based Framework for Data-Driven Clothing 2020 / SCA
- Motion Guided Deep Dynamic 3D Garments 2022 / SIGGRAPH Asia
Keywords
This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →