Search papers, labs, and topics across Lattice.
This paper introduces CoGeoAD, a unified framework for zero-shot 3D anomaly detection that effectively integrates 2D color images and 3D geometric structures through a novel Data-Driven Multi-View Attention mechanism and a Multi-Stage Color-Geometric Fusion module. The approach addresses the limitations of existing methods by enabling the detection of both surface and structural defects in a cohesive manner, leveraging pixel-aligned paired multi-view images. Experimental results on the MVTec3D-AD and Eyecandies benchmarks show that CoGeoAD outperforms state-of-the-art methods, significantly enhancing detection capabilities in complex industrial settings.
State-of-the-art performance in zero-shot 3D anomaly detection is achieved by seamlessly fusing color and geometric features, revealing structural and textural anomalies that others miss.
Zero-shot 3D anomaly detection is essential for industrial quality inspection, where labeled anomaly samples are scarce. Meanwhile, existing methods lack an effective mechanism to fuse complementary 2D color images with 3D geometric structures, limiting their ability to detect both surface and structural defects in a unified framework. To address these issues, we propose CoGeoAD, a unified CLIP-based framework that fuses color and geometric features by constructing pixel-aligned paired multi-view images. The framework introduces a Data-Driven Multi-View Attention (MVA) mechanism to adaptively aggregate 3D features and a Multi-Stage Color-Geometric Fusion (MS-CGF) module to hierarchically integrate multi-level features from both modalities. Extensive experiments on the MVTec3D-AD and Eyecandies benchmarks demonstrate that CoGeoAD achieves state-of-the-art performance, effectively capturing both structural and textural anomalies in complex industrial scenarios. our source code is available at https://github.com/kingdomShu/CoGeoAD.
