Graph Based Texture Pattern Classification

Textures in 3D meshes represent intrinsic surface properties and are essential for various applications, including retrieval, segmentation, and classification. However, it is distinct from other types of 3D object analysis. The primary objective is to capture the surface variations induced by multiple textures. While numerous classical approaches are published in the literature, only a few work directly on 3D meshes. Given the versatility of graph representations, we propose a graph learning-based approach for classifying the texture of each facet in a 3D mesh. First, a three-dimensional mesh is transformed into a graph structure in which every node is a facet of a given mesh. Further, each facet is described by a feature vector computed utilizing the neighboring facets within a radius and their geometric properties. The graph structure is then fed into a graph neural network, classifying each node as a texture or non-textured class. The proposed technique has been validated using texture patterns from SHREC'18 and demonstrated positive performance.