GAPS: Geometry-Aware, Physics-Based, Self-Supervised Neural Garment Draping

TL;DR

GAPS tackles realistic garment draping by enforcing collision-aware inextensibility through geometry-based constraints and introducing a geometry-aware skinning scheme. The method integrates a covariance-based local inextensibility model with a Gaussian-RBF body-participation mechanism within an unsupervised GRU framework, combining physics-based losses with a novel inextensibility term. Empirical results on AMASS show improved isometry, reduced collisions, and applicability to tight and loose garments across diverse bodies, outperforming prior self-supervised and body-specific approaches. The work offers a scalable, real-time-friendly alternative that mitigates post-processing and broadens generalization to various garment topologies.

Abstract

Recent neural, physics-based modeling of garment deformations allows faster and visually aesthetic results as opposed to the existing methods. Material-specific parameters are used by the formulation to control the garment inextensibility. This delivers unrealistic results with physically implausible stretching. Oftentimes, the draped garment is pushed inside the body which is either corrected by an expensive post-processing, thus adding to further inconsistent stretching; or by deploying a separate training regime for each body type, restricting its scalability. Additionally, the flawed skinning process deployed by existing methods produces incorrect results on loose garments. In this paper, we introduce a geometrical constraint to the existing formulation that is collision-aware and imposes garment inextensibility wherever possible. Thus, we obtain realistic results where draped clothes stretch only while covering bigger body regions. Furthermore, we propose a geometry-aware garment skinning method by defining a body-garment closeness measure which works for all garment types, especially the loose ones.

Figures (9)

• Figure 1: SNUGSantesteban22 and NCSBertiche22 perform self-supervised draping of garments. However, they are prone to unrealistic stretching arising due to their efforts to prevent garment-body collisions. To overcome this issue, we propose GAPS which performs geometrically aware mechanism to drape a garment by stretching only if necessary. This controls body-garment collisions automatically without any post-processing Santesteban22 or restrictive measures Bertiche22. Furthermore, our geometry-aware skinning allows better handling of loose garments.
• Figure 2: Body-participation computation for garment vertices close (in blue) or far (in red) to the body.
• Figure 3: Overview of our method. We added RBF based Body Participation module on Santesteban22 to estimate blend weights based on relative garment-body geometry.
• Figure 4: T-shirt, Dress, Skirt, and Gown draped using GAPS.
• Figure 5: Results on AMASS sequence 86_07 with Dress. GAPS shows the best performance with minimal collision and realistic stretching (indicated in orange).