CloSE: A Compact Shape- and Orientation-Agnostic Cloth State Representation

TL;DR

This work introduces CloSE, a compact, shape- and orientation-agnostic cloth state descriptor built from the novel dGLI disk, a circular arrangement of border-edge dGLI values. CloSE encodes cloth state via ordered corner angles and fold endpoints with orientation, enabling continuous interpolation between states and reconstruction of cloth borders. The authors demonstrate two key applications—semantic labeling and planning—achieving high fold-detection accuracy on multiple datasets and enabling reasoning about folds without learning. Overall, CloSE provides a practical, generalizable framework for reasoning about non-rigid cloth states, with potential impact on robot manipulation and planning in cluttered, deformable-object environments.

Abstract

Cloth manipulation is a difficult problem mainly because of the non-rigid nature of cloth, which makes a good representation of deformation essential. We present a new representation for the deformation-state of clothes. First, we propose the dGLI disk representation, based on topological indices computed for segments on the edges of the cloth mesh border that are arranged on a circular grid. The heat-map of the dGLI disk uncovers patterns that correspond to features of the cloth state that are consistent for different shapes, sizes of positions of the cloth, like the corners and the fold locations. We then abstract these important features from the dGLI disk onto a circle, calling it the Cloth StatE representation (CloSE). This representation is compact, continuous, and general for different shapes. Finally, we show the strengths of this representation in two relevant applications: semantic labeling and high- and low-level planning. The code, the dataset and the video can be accessed from : https://jaykamat99.github.io/close-representation

Figures (9)

• Figure 1: Example of CloSE derivation: From the border, we compute the dGLI disk where each petal corresponds to a corner. The difference between dGLI disks for the start and end frames allows to extract the fold information, so as to obtain the final CloSE descriptor that is shown graphically and numerically at the bottom of the figure.
• Figure 2: Behavior of dGLI in case of planar segments
• Figure 3: a) A square napkin mesh. b) The corresponding original dGLI matrix as in Franco-dGLI and c) Our proposed arrangement in the dGLI disk
• Figure 4: The proposed dGLI disk mapping of the values of dGLI$(i,j)$
• Figure 5: Column a) shows the border of the cloth mesh, b) the dGLI disk difference, c) the clustered selected points from the dGLI disk difference, d) shows the resulting fitted curve and the computed fold coordinates, printed over the dGLI disk difference, and e) shows the reconstructed mesh from the initial cloth border using the CloSE representation.