Detecting 3D Texture on Cultural Heritage Artifacts

Abstract

Textures in 3D meshes represent intrinsic surface properties and are essential for numerous applications, such as retrieval, segmentation, and classification. The computer vision approaches commonly used in the cultural heritage domain are retrieval and classification. Mainly, these two approaches consider an input 3D mesh as a whole, derive features of global shape, and use them to classify or retrieve. In contrast, texture classification requires objects to be classified or retrieved based on their textures, not their shapes. Most existing techniques convert 3D meshes to other domains, while only a few are applied directly to 3D mesh. The objective is to develop an algorithm that captures the surface variations induced by textures. This paper proposes an approach for texture classification directly applied to the 3D mesh to classify the surface into texture and non-texture regions. We employ a hybrid method in which classical features describe each facet locally, and these features are then fed into a deep transformer for binary classification. The proposed technique has been validated using SHREC’18 texture patterns, and the results demonstrate the proposed approach’s effectiveness.

Keywords 3D Texture, Binary classification, 3D mesh , SHREC’18.