Introduction of new dataset for unsupervised fabric defect detection This dataset aims to provide a color dataset with real industrial fabric defect gathered in a visiting machine with several industrial cameras. It has been designed with the same nomenclature as MVTEC AD dataset (https://www.mvtec.com/company/research/datasets/mvtec-ad) for unsupervised anomaly detection.
The dataset can be downloaded in google drive with this link : LINK
This dataset is designed for unsupervised anomaly detection task but can also be used for domain-generalization approach. The nomenclature is designed as :
As in any unsupervised training, train data are defect-free. Defective samples are only in the test set.
Exemple of defect segmentation obtained with our knowledge distillation-based method
List of articles related to the subject of textile defect detection
MixedTeacher : Knowledge Distillation for fast inference textural anomaly detection (https://arxiv.org/abs/2306.09859) (https://github.com/SimonThomine/MixedTeacher)
FABLE : Fabric Anomaly Detection Automation Process (https://arxiv.org/abs/2306.10089)
Exploring Dual Model Knowledge Distillation for Anomaly Detection (https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4493018)
Distillation-based fabric anomaly detection (https://journals.sagepub.com/doi/abs/10.1177/00405175231206820)(https://arxiv.org/abs/2401.02287) (https://github.com/SimonThomine/DBFAD)
CSE: Surface Anomaly Detection with Contrastively Selected Embedding (https://arxiv.org/pdf/2403.01859.pdf) (https://github.com/SimonThomine/CSE)
Simon Thomine 1, PhD student - @SimonThomine - simon.thomine@utt.fr
Hichem Snoussi 1, Full Professor
1 University of Technology of Troyes, France
If you use this dataset, please cite
@inproceedings{Thomine_2023_Knowledge,
author = {Thomine, Simon and Snoussi, Hichem},
title = {Distillation-based fabric anomaly detection},
booktitle = {Textile Research Journal},
month = {August},
year = {2023}
}
This project is under the MIT license MIT.