/AutoEncoder_vs_MetricLearning

AutoEncoder vs Metric Learning for Anomaly Detection

Primary LanguageJupyter NotebookMIT LicenseMIT

1.AutoEncoder_vs_MetricLearning

  • AutoEncoder vs Metric Learning for Anomaly Detection
  • Dataset:MVTEC-AD

Click here for details(Japanese).

2.Simple method

Picture L2-SoftmaxLoss(AUC) Simple method(AUC)
Cable 0.72 0.91
Leather 0.91 0.99
Carpet 0.71 0.8
Capsule 0.57 0.87

Click here for details(Japanese).

3.Self Supervised(class3)

fig fig2

Click here for details(Japanese).