← Back ICRA 2026

3D-ADAM: A Dataset for 3D Anomaly Detection in Additive Manufacturing

Paul Matthew McHard, Florent P. Audonnet, Oliver Summerell, Sebastian Andraos, Paul Henderson, Gerardo Aragon-Camarasa

PDF

AI summary

Key figure (auto-extracted from paper)

Current state-of-the-art 3D anomaly detection models struggle to generalize to realistic industrial conditions, as demonstrated by their significant underperformance on the newly introduced 3D-ADAM dataset.

3D anomaly detection additive manufacturing industrial dataset unsupervised learning defect detection RGB-D data

Problem

Existing 3D anomaly detection datasets lack the scale, defect diversity, and real-world industrial variability required to develop models capable of reliable deployment in additive manufacturing.

Approach

The authors compiled a large-scale, multi-sensor dataset of 14,120 high-resolution RGB and 3D scans of 217 additive manufacturing parts with extensive defect annotations, then benchmarked leading unsupervised anomaly detection models against this realistic industrial benchmark.

Key results

Release of 3D-ADAM, a large-scale multi-sensor dataset with 14,120 scans and 27,346 defect annotations
Comprehensive coverage of 12 defect classes and 16 machine element features across 28 part categories
Benchmarking reveals significant performance gaps in current state-of-the-art unsupervised anomaly detection models
Expert validation confirms high annotation accuracy and direct industrial applicability

Why it matters

This dataset provides a critical industrial benchmark to bridge the gap between academic research and real-world manufacturing, enabling the development of robust 3D anomaly detection for automated quality control and robotic defect correction.

Abstract

Surface defects are a primary source of yield loss in manufacturing, yet existing anomaly detection methods often fail in real-world deployment due to limited and unrepresen- tative datasets. To overcome this, we introduce 3D-ADAM, a 3D Anomaly Detection in Additive Manufacturing dataset, that is the first large-scale, industry-relevant dataset for RGB+3D surface defect detection in additive manufacturing. 3D-ADAM comprises 14,120 high-resolution scans of 217 unique parts, captured with four industrial depth sensors, and includes 27,346 annotated defects across 12 categories along with 27,346 annotations of machine element features in 16 classes. 3D- ADAM is captured in a real industrial environment and as such reflects real production conditions, including variations in part placement, sensor positioning, lighting, and partial occlusion. Benchmarking state-of-the-art models demonstrates that 3D- ADAM presents substantial challenges beyond existing datasets. Validation through expert labelling surveys with industry part- ners further confirms its industrial relevance. By providing this benchmark, 3D-ADAM establishes a foundation for advancing robust 3D anomaly detection capable of meeting manufacturing demands. We provide our dataset for accessibility at: https: //huggingface.co/datasets/pmchard/3D-ADAM

Index terms

Computer Vision for Manufacturing Additive Manufacturing Data Sets for Robotic Vision