Title: CloSE: A Geometric Shape-Agnostic Cloth State Representation

Abstract:

Manipulating cloth is a significant challenge, primarily due to its non-rigid properties, which necessitates a robust method for representing deformation. To address this, we introduce a novel approach for capturing the deformation state of garments. Our method begins with the dGLI disk representation, which utilizes topological indices calculated from edge segments along the cloth’s perimeter, organized onto a circular grid. The resulting heat map of the dGLI disk reveals patterns associated with specific cloth state features that remain consistent regardless of the garment’s shape, size, or orientation. We then distill these key characteristics from the dGLI disk into a circular format, termed the Cloth StatE (CloSE) representation. CloSE is characterized by its compactness, continuity, and generalizability across various geometric forms. We demonstrate that this representation can accurately forecast fold locations across multiple simulation-based clothing datasets. Furthermore, we highlight the utility of CloSE in two key applications: semantic labeling and both high- and low-level planning. The associated code and dataset are available at: https://close-representation.github.io/

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC