Skip to content

A PyTorch-based framework for unsupervised anomaly detection on the MVTec AD dataset. Implements PaDiM, PatchCore, FastFlow, and AutoEncoder models with utilities for dataset preparation, training, evaluation, and reporting.

Notifications You must be signed in to change notification settings

taha-kms/defect-detection

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

48 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Defect Detection on MVTec AD

This repository provides a modular framework for unsupervised anomaly detection and localization on the MVTec Anomaly Detection (AD) dataset.

It implements several popular models for visual anomaly detection:

  • Autoencoder (AE) – convolutional autoencoder with SSIM + MSE loss
  • PaDiM – patch distribution modeling with multivariate Gaussians
  • PatchCore – memory-bank nearest-neighbor matching on patch embeddings
  • FastFlow – normalizing-flow–based density estimation on features

Utilities are included for data preparation, training, evaluation, and reporting, with metrics like Image-level AUROC, Pixel-level AUROC, AUPRC, and PRO.

About

A PyTorch-based framework for unsupervised anomaly detection on the MVTec AD dataset. Implements PaDiM, PatchCore, FastFlow, and AutoEncoder models with utilities for dataset preparation, training, evaluation, and reporting.

Topics

Resources

Stars

Watchers

Forks