ISSTAD

ISSTAD: Incremental Self-Supervised Learning Based on Transformer for Anomaly Detection and Localization

Introduction

This repository contains the source code for ISSTAD implemented on the MVTec AD dataset and the MVTec LOCO AD dataset.

Get Started

Datasets

https://www.mvtec.com/company/research/datasets/mvtec-ad https://www.mvtec.com/company/research/datasets/mvtec-loco

Pre-trained MAE model

Kindly obtain the pre-trained MAE model from the provided link.
https://dl.fbaipublicfiles.com/mae/visualize/mae_visualize_vit_large.pth

Environment

python==3.9.13
matplotlib==3.6.0
numpy==1.23.3
opencv-python==4.6.0.66
pandas==1.5.1
pillow==9.2.0
scikit-learn==1.1.2
scipy==1.9.1
six==1.16.0
timm==0.3.2
torch==1.12.1+cu116
tqdm==4.64.1

The code is executable on Windows systems, and if running on Linux, it requires execution on a disk with an NTFS file system. Otherwise, the results may degrade, especially for the localization result on the MVTec AD dataset.

Run

MVTec AD dataset

python main_mvtec_ad.py

MVTec LOCO AD dataset

python main_mvtec_loco_ad.py