LudovicoL / PaDiM

PaDiM - Anomaly Detection Localization

Extended implementation of PaDiM.

Original paper: PaDiM: a Patch Distribution Modeling Framework for Anomaly Detection and Localization

Datasets

• MVTec AD: Download from MVTec website
• AITEX: Download from AITEX website
• BTAD: Download from AViReS Laboratory website

How to use

To train the model, use:

python train.py

To evaluate the model:

python evaluate.py

The option -d permits to choose the dataset (aitex, mvtec, btad). In the train file, the datasets will be automatically downloaded.

Minimum requirements

Python 3.9 with PyTorch 1.9.0. Use the file environment.yml for the conda environment.

Reference

[1] Thomas Defard, Aleksandr Setkov, Angelique Loesch, Romaric Audigier. PaDiM: a Patch Distribution Modeling Framework for Anomaly Detection and Localization. https://arxiv.org/pdf/2011.08785

[2] https://github.com/xiahaifeng1995/PaDiM-Anomaly-Detection-Localization-master