Dataset · Voxel51
MVTec AD industrial defect dataset in FiftyOne format
Voxel51/mvtec-ad
Well-known benchmark for anomaly detection and manufacturing defects (5,354 images) adapted for FiftyOne.
What's inside
Contains 5,354 images from the original MVTec AD dataset with annotations for classification and segmentation tasks. Data is converted for convenient and fast loading directly via the FiftyOne Python library.
What you can do with it
- Training and validating quality control and manufacturing defect detection models
- Testing anomaly detection algorithms in computer vision
- Rapid prototyping and visual error analysis in the FiftyOne environment
- Tasks
- Image classification, Image segmentation
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 76.7K
- Likes
- 18
How to load it
from datasets import load_dataset
ds = load_dataset("Voxel51/mvtec-ad")