datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
mvtec-ad
Dataset Card for MVTec AD
This dataset originates from MVTec but is provided in a different format. You can easily load it using FiftyOne
The total number of samples remains the same as the original: 5,354.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
import fiftyone.utils.huggingface as fouh
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset =… See the full description on the dataset page: https://huggingface.co/datasets/Voxel51/mvtec-ad.mvtec_anomaly_detectionmvtec-ad
MVTec AD - Anomaly Detection Dataset
This repository mirrors the MVTec AD dataset for CI/CD and non-commercial research workflows.
License
The data is released under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0). It is not allowed to use the dataset for commercial purposes.
Attribution
If you use the dataset in scientific work, please cite the original authors:
Paul Bergmann, Kilian Batzner… See the full description on the dataset page: https://huggingface.co/datasets/foersben/mvtec-ad.mvtec_locomvtech_anomaly_detection
MVTec Anomaly Detection Dataset (MVTec AD)
Dataset description
The MVTec Anomaly Detection (MVTec AD) dataset is a large-scale real-world dataset for unsupervised anomaly detection in industrial inspection scenarios.It contains high-resolution images of multiple object and texture categories, including normal samples and various defect types with pixel-level ground truth masks.
Official website:https://www.mvtec.com/company/research/datasets/mvtec-ad
Dataset… See the full description on the dataset page: https://huggingface.co/datasets/micguida1/mvtech_anomaly_detection.mvtec-ad
Dataset Card for MVTec AD
This dataset originates from MVTec but is provided in a different format. You can easily load it using FiftyOne
The total number of samples remains the same as the original: 5,354.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
import fiftyone.utils.huggingface as fouh
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset… See the full description on the dataset page: https://huggingface.co/datasets/OctorVX/mvtec-ad.MVTec-4KHigh-resolution industrial image anomaly detection dataset MVTec-4K.For more information, see HiAD.
Download
huggingface-cli download --repo-type dataset XimiaoZhang/MVTec-4K --local-dir MVTec-4K --resume-download
mvtec_all_objects_splitmvtech_anomaly_detectionmvtec-ad
Dataset Card for MVTec AD
This dataset originates from MVTec but is provided in a different format. You can easily load it using FiftyOne
The total number of samples remains the same as the original: 5,354.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
import fiftyone.utils.huggingface as fouh
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset… See the full description on the dataset page: https://huggingface.co/datasets/box0602/mvtec-ad.mirage_mvtec_visaMVTecAD2_FO
Dataset Card for mvtecad2
This is a FiftyOne dataset with 3914 samples.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
from fiftyone.utils.huggingface import load_from_hub
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset = load_from_hub("pjramg/MVTecAD2_FO")
# Launch the App
session = fo.launch_app(dataset)
Dataset Details… See the full description on the dataset page: https://huggingface.co/datasets/pjramg/MVTecAD2_FO.mvtec_capsule
MVTec Capsule Category
Dataset Labels
{0: "normal", 1: "abnormal"}
Number of Images
{'train': 219, 'test': 132}
How to Use
Install datasets:
pip install datasets
Load the dataset:
from datasets import load_dataset
ds = load_dataset("alexsu52/mvtec_capsule")
example = ds['train'][0]
MVTEC Dataset Page
https://www.mvtec.com/company/research/datasets/mvtec-ad
Citation
Paul Bergmann, Kilian Batzner, Michael Fauser… See the full description on the dataset page: https://huggingface.co/datasets/alexsu52/mvtec_capsule.mvtec-ad-metal-nut
MVTec AD - Metal Nut Dataset
Metal nut subset from MVTec Anomaly Detection dataset.
Defect Types
bent, color, flip, scratch
AU-ROC: 99.7% (EfficientAD medium)
synthetic-mvtec-ad-defect-detection
Synthetic MVTec AD – Defect Detection Dataset by AnywayLabs.ai
Need a custom synthetic dataset for your own defect detection use case?
This dataset is an open-source sample of our synthetic data generation work at AnywayLabs.
If you're working on:
industrial defect detection
visual inspection
supervised anomaly detection
hard-to-collect defect classes
synthetic data for computer vision training
You can request a custom synthetic dataset here, or email:… See the full description on the dataset page: https://huggingface.co/datasets/anywaylabs/synthetic-mvtec-ad-defect-detection.MVTecADmvtec-ad
Dataset Card for MVTec AD
This dataset originates from MVTec but is provided in a different format. You can easily load it using FiftyOne
The total number of samples remains the same as the original: 5,354.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
import fiftyone.utils.huggingface as fouh
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset… See the full description on the dataset page: https://huggingface.co/datasets/pmrozo/mvtec-ad.potato_mvtec3dmvtec-ad
Dataset Card for MVTec AD
This dataset originates from MVTec but is provided in a different format. You can easily load it using FiftyOne
The total number of samples remains the same as the original: 5,354.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
import fiftyone.utils.huggingface as fouh
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset… See the full description on the dataset page: https://huggingface.co/datasets/rfvbbjui/mvtec-ad.MVTec-FSThis dataset is an extended version of the MVTec-AD anomaly detection dataset, specifically designed for anomaly classification tasks.
It is built upon the MVREC (AAAI 2025) framework. We have reorganized the format, and we would like to acknowledge the excellent foundational work that preceded this.
Download Command:
huggingface-cli download --repo-type dataset XimiaoZhang/MVTec-FS --local-dir ./ --resume-download
mvtecmvtec-ad-cable
MVTec AD - Cable Dataset
This is the Cable subset from the MVTec Anomaly Detection dataset.
Dataset Structure
cable/
├── train/
│ └── good/ # 224 normal training images
└── test/
├── good/ # 58 normal test images
├── bent_wire/ # 14 images
├── cable_swap/ # 14 images
├── combined/ # 13 images
├── cut_inner_insulation/ # 16 images
├── cut_outer_insulation/ # 12 images
├── missing_cable/ # 14 images
├──… See the full description on the dataset page: https://huggingface.co/datasets/MSherbinii/mvtec-ad-cable.mvtec_yolo
MVTec D2S Dataset
This dataset is a modification of the D2S set to the COCO format.
The Annotations for the test set are not public and the results can be requested per mail if in the correct format.
The Augmentation set (10k images) is split into Train and Val atm (70-30) but we should maybe change this as the val set is important since we don't have the annotations of the test set
mvtec-adOriginal dataset:
@inproceedings{bergmann2019mvtec,
title={MVTec AD--A comprehensive real-world dataset for unsupervised anomaly detection},
author={Bergmann, Paul and Fauser, Michael and Sattlegger, David and Steger, Carsten},
booktitle={Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},
pages={9592--9600},
year={2019}
}
mvtecad-woodMVTec-2KHigh-resolution industrial image anomaly detection dataset MVTec-2K.For more information, see HiAD.
Download
huggingface-cli download --repo-type dataset XimiaoZhang/MVTec-2K --local-dir MVTec-2K --resume-download
mvtec-ad
Dataset Card for MVTec AD
This dataset originates from MVTec but is provided in a different format. You can easily load it using FiftyOne
The total number of samples remains the same as the original: 5,354.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
import fiftyone.utils.huggingface as fouh
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset… See the full description on the dataset page: https://huggingface.co/datasets/introvoyz041/mvtec-ad.mvtec_anomaly_detectionmvtec_mapped
Mapped MVTec Sets
This Repo contains different test sets which are subsets of the actual MVTec Set but mapped to our labels and filtered.
mvtec_anomaly_classification