IEC3D-AD: A 3D Dataset of Industrial Equipment Components for Unsupervised Point Cloud Anomaly Detection

TL;DR

This work tackles the challenge of 3D anomaly detection in industrial manufacturing by introducing IEC3D-AD, a real-production-line, high-resolution 360-degree point cloud dataset with fine-grained defect annotations. It proposes GMANet, an unsupervised 3D-AD paradigm that synthesizes abnormal samples via geometric morphological analysis (SPCG) and learns robust normal-abnormal feature separation through spatial discrepancy optimization (SDO) across dual encoders. The dataset offers expansive category coverage, high data fidelity, and small defect proportions to stress-test algorithms, while the benchmark and analyses demonstrate strong improvements over state-of-the-art methods on IEC3D-AD and competitive performance on related datasets. This work thus provides a practical, scalable foundation for industrial 3D anomaly detection with direct implications for reliability and safety in manufacturing environments.

Abstract

3D anomaly detection (3D-AD) plays a critical role in industrial manufacturing, particularly in ensuring the reliability and safety of core equipment components. Although existing 3D datasets like Real3D-AD and MVTec 3D-AD offer broad application support, they fall short in capturing the complexities and subtle defects found in real industrial environments. This limitation hampers precise anomaly detection research, especially for industrial equipment components (IEC) such as bearings, rings, and bolts. To address this challenge, we have developed a point cloud anomaly detection dataset (IEC3D-AD) specific to real industrial scenarios. This dataset is directly collected from actual production lines, ensuring high fidelity and relevance. Compared to existing datasets, IEC3D-AD features significantly improved point cloud resolution and defect annotation granularity, facilitating more demanding anomaly detection tasks. Furthermore, inspired by generative 2D-AD methods, we introduce a novel 3D-AD paradigm (GMANet) on IEC3D-AD. This paradigm generates synthetic point cloud samples based on geometric morphological analysis, then reduces the margin and increases the overlap between normal and abnormal point-level features through spatial discrepancy optimization. Extensive experiments demonstrate the effectiveness of our method on both IEC3D-AD and other datasets.

Figures (8)

• Figure 1: IEC3D-AD possesses a small defect ratio while ensuring data authenticity and spatial coverage. It can be seen that defects in real industrial scenarios are significantly different from others.
• Figure 2: The example of IEC3D-AD for each category.
• Figure 3: (a) Acquisition system. (b) The pipeline of calibration process. The relevant parameters generated through the calibration process can help us complete the coarse and fine registration tasks of point cloud data.
• Figure 4: Comparison of abnormal proportions distribution and point cloud number of different datasets.
• Figure 5: The pipeline of GMANet. During the training phase, we conduct a novel synthetic point cloud generation (SPCG) module based on geometric morphological analysis (GMA) to generate synthetic abnormal samples. Then, our spatial discrepancy optimization (SDO) module simultaneously reduces the margin between the normal and synthetic abnormal feature distribution extracted from different encoders and expands the overlap of the feature distribution of similar features between different encoders. During testing phase, we conduct an anomaly score calculation function to detect abnormal points.