/AnomalyDetection

A system capable of anomaly detection for two distinct products from the MVTec Anomaly Detection dataset.

Primary LanguagePythonApache License 2.0Apache-2.0

Anomaly Detection System for MVTec AD

This is a system capable of anomaly detection for two distinct products from the MVTec Anomaly Detection dataset, screw and metal nut. alt text The whole pipeline is implemented via pipeline.py.

Methods

The machine learning model is based on EfficientAD. https://arxiv.org/abs/2303.14535. The code is built based on https://github.com/nelson1425/EfficientAD.

By using Lightweight Student–Teacher + Autoencoder architecture for anomalies detection and Patch description networks (PDN) for feature extraction. The model enables a fast handling of anomalies with low error rate, making it a perfect choise for abnomaly detection in manufacturing industry.

Challenges

  1. Types and possible locations of defects are unknown --> Use Student–Teacher structure so that it perform well even trained only on normal images.

  2. Industrial settings requires strict runtime limits --> Reduce computational cost by drastically reducing depth for feature extractor, performing down-sampling early, and using light-weight S-T structure.

  3. Violations of logical constraints regarding the position, size, arrangement, etc. --> Use Autoencoder to detect logical anomalies.

Results

alt text

Mean anomaly detection AU-ROC percentages:

Product Model AU-ROC
screw EfficientAD-S 96.8
screw EfficientAD-M 97.4
metal nut EfficientAD-S 98.7
metal nut EfficientAD-M 99.3

Computational efficiency: Latency

Model GPU Latency
EfficientAD-S Quadro RTX 6000 4.0 ms
EfficientAD-M Quadro RTX 6000 5.9 ms

Setup

Packages

Python==3.10
numpy==1.18.5
torch==1.13.0
torchvision==0.14.0
scikit-learn==1.2.2
tifffile==2021.7.30
tqdm==4.56.0
Pillow==7.0.0
scipy==1.7.1
tabulate==0.8.7

Dataset

Download dataset (if you already have downloaded then set path to dataset (--mvtec_ad_path) when calling efficientad.py).

mkdir mvtec_anomaly_detection
cd mvtec_anomaly_detection
wget https://www.mydrive.ch/shares/38536/3830184030e49fe74747669442f0f282/download/420938113-1629952094/mvtec_anomaly_detection.tar.xz
tar -xvf mvtec_anomaly_detection.tar.xz
cd ..

Usage

Training and inference for screw and metal nut:

python mvtec_ad_training/efficientad_2objects.py --dataset mvtec_ad

Training with EfficientAD-M for two kinds of images:

python mvtec_ad_training/efficientad_2objects.py --model_size medium --weights models/teacher_medium.pth --dataset mvtec_ad

Evaluation with Mvtec evaluation code:

python mvtec_ad_evaluation/evaluate_experiment.py --anomaly_maps_dir './output/4/anomaly_maps/mvtec_ad/' --output_dir './output/4/metrics/mvtec_ad/' --evaluated_objects screw

Anomaly detection pipeline for one image:

python pipeline.py --sample_path './mvtec_anomaly_detection/metal_nut/test/scratch/000.png'