Flatten Anything: Unsupervised Neural Surface Parameterization

TL;DR

This paper introduces the Flatten Anything Model (FAM), an unsupervised neural architecture to achieve global free-boundary surface parameterization via learning point-wise mappings between 3D points on the target geometric surface and adaptively-deformed UV coordinates within the 2D parameter domain.

Abstract

Surface parameterization plays an essential role in numerous computer graphics and geometry processing applications. Traditional parameterization approaches are designed for high-quality meshes laboriously created by specialized 3D modelers, thus unable to meet the processing demand for the current explosion of ordinary 3D data. Moreover, their working mechanisms are typically restricted to certain simple topologies, thus relying on cumbersome manual efforts (e.g., surface cutting, part segmentation) for pre-processing. In this paper, we introduce the Flatten Anything Model (FAM), an unsupervised neural architecture to achieve global free-boundary surface parameterization via learning point-wise mappings between 3D points on the target geometric surface and adaptively-deformed UV coordinates within the 2D parameter domain. To mimic the actual physical procedures, we ingeniously construct geometrically-interpretable sub-networks with specific functionalities of surface cutting, UV deforming, unwrapping, and wrapping, which are assembled into a bi-directional cycle mapping framework. Compared with previous methods, our FAM directly operates on discrete surface points without utilizing connectivity information, thus significantly reducing the strict requirements for mesh quality and even applicable to unstructured point cloud data. More importantly, our FAM is fully-automated without the need for pre-cutting and can deal with highly-complex topologies, since its learning process adaptively finds reasonable cutting seams and UV boundaries. Extensive experiments demonstrate the universality, superiority, and inspiring potential of our proposed neural surface parameterization paradigm. Our code is available at https://github.com/keeganhk/FlattenAnything.

Figures (13)

• Figure 1: Flatten Anything Model (FAM) for neural surface parameterization: (a) input 3D models; (b) learned UV coordinates; (c) texture mappings; (d) learned cutting seams.
• Figure 2: Illustration of bi-directional cycle mapping, composed of (a) 2D$\rightarrow$3D$\rightarrow$2D cycle mapping branch, and (b) 3D$\rightarrow$2D$\rightarrow$3D cycle mapping branch. Modules with the same color share network parameters. (c) shows the learned cutting seams. (d) shows the checker-image texture mapping.
• Figure 3: Comparison of UV unwrapping and texture mapping results on different (a) open surface models produced by (b) our FAM and (c) SLIM, where the 2D UV coordinates are color-coded by ground-truth point-wise normals to facilitate visualization.
• Figure 4: Display of surface parameterization results produced by our FAM. (a) input 3D models; (b) learned UV coordinates; (c) texture mappings; (d) learned cutting seams.
• Figure 5: Point cloud parameterization achieved by our FAM (left) and FBCP-PC (right).