서론

목차

  1. Introduction & Related Work
    1. background Research
    2. Recent Research
    3. Propose
  2. Abstract
  3. Introduction
  4. Related work
  5. DRÆM
    1. 3.1 Reconstructive subnetwork
    2. 3.2. Discriminative subnetwork
    3. 3.3. Simulated anomaly generation
    4. 3.4. Surface anomaly localization and detection
  6. Experiments
    1. Datasets
      1. MVTec(15 classes)
      2. DTD(Describable Textures Dataset)
    2. Setting
    3. Surface Anomaly Detection
    4. Anomaly localization
    5. Ablation Study
    6. Comparison with supervised methods

Introduction & Related Work

background Research

Recent surface anomaly detection methods rely on generative models to accurately reconstruct the normal areas and to fail on anomalies.

Recent Research

Reconstructive methods, such as Autoencoders [5, 1, 2, 26] and GANs [24, 23], have been extensively explored since they enable learning of a powerful reconstruction subspace, using only anomaly-free images.

These methods are trained only on anomaly-free images, and often require hand-crafted post-processing steps to localize the anoma- lies, which prohibits optimizing the feature extraction for maximal detection capability.