3D-ADAM: 3D Anomaly Detection in Additive Manufacturing with 1K-10K Records

3D-ADAM contains between 1,000 and 10,000 records for anomaly detection in additive manufacturing, published by pmchard in 2025. The data includes raw 3D part-instances and a single-camera subset captured via MechMind-Nano sensors for unsupervised training tasks.

Use Cases

• Unsupervised anomaly detection using the MechMind-Nano sensor subset
• Anomaly localization on 3D printed part-instances
• Segmentation of manufacturing defects using imagefolder formatted data

Strengths

• 1,000 to 10,000 records
• Expert-curated for the anomalib library
• Includes raw part-instance data for custom processing

Limitations

• Restricted to additive manufacturing domain
• CC BY-NC-SA 4.0 license prohibits commercial use

Provenance

Source

pmchard, ArXiv:2507.07838

Collection Method

Sensor-based capture using MechMind-Nano cameras in an additive manufacturing environment.

Freshness

Last updated October 2025.

Geography

United States