Skip to content

Dataset · foersben

MVTec AD Benchmark Mirror for Photo Anomaly Detection

foersben/mvtec-ad

Well-known dataset for detecting defects in images (non-commercial use only).

What's inside

The repository is a mirror of the popular MVTec AD dataset (10k to 100k images) for classification and object detection tasks, prepared for CI/CD and research workflows. The data is distributed strictly under the CC BY-NC-SA 4.0 non-commercial license with the requirement to cite original authors.

What you can do with it

  • Training and validating computer vision models for manufacturing defect detection
  • Comparing anomaly detection algorithms on a standard benchmark
  • Automated testing of visual inspection pipelines in non-commercial research
Tasks
Image classification, Object detection
Size
10K–100K rows
Downloads this month
55.4K
Likes
8

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("foersben/mvtec-ad")