Multi-View Industrial Parts with Anomaly Masks

A 360-degree all-around view collection of 7 industrial parts, captured with one axis of rotation. It includes a training set of all-around images and test sets with ground truth anomaly masks. The dataset was created by author kentaito321 and was last updated in March 2026.

Use Cases

• Train anomaly detection models using the ground truth anomaly masks provided in the test sets.
• Develop multi-view 3D reconstruction algorithms from the 360-degree all-around image sequences.
• Benchmark robustness of visual inspection systems across different viewpoints of the 7 industrial parts.
• Research free-viewpoint synthesis for industrial inspection using the structured rotation axis data.

Strengths

• Contains ground truth anomaly masks for evaluation.
• Features a structured 360-degree view with one axis of rotation for each part.
• Dataset was updated in March 2026, indicating recent maintenance.

Limitations

• Limited to only 7 distinct industrial parts, which may restrict model generalization.
• Specific row counts, image resolutions, and file formats are unknown from the description.

Provenance

Source

huggingface dataset by author kentaito321, associated with the paper '3D Gaussian Reference Parts for Robust Free-Viewpoint Visual Inspection'.

Collection Method

Images taken from a 360-degree all-around view with one axis of rotation.

Freshness

Last updated 2026-03-20.