Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Reconstruction-Based Anomaly Localization via Knowledge-Informed Self-Training

Cheng Qian, Xiaoxian Lao, Chunguang Li

TL;DR

A novel reconstruction-based method named knowledge-informed self-training (KIST) which integrates knowledge into a reconstruction model through self-training and exploits knowledge to yield pixel-level pseudolabels of the anomalous samples.

Abstract

Anomaly localization, which involves localizing anomalous regions within images, is a significant industrial task. Reconstruction-based methods are widely adopted for anomaly localization because of their low complexity and high interpretability. Most existing reconstruction-based methods only use normal samples to construct model. If anomalous samples are appropriately utilized in the process of anomaly localization, the localization performance can be improved. However, usually only weakly labeled anomalous samples are available, which limits the improvement. In many cases, we can obtain some knowledge of anomalies summarized by domain experts. Taking advantage of such knowledge can help us better utilize the anomalous samples and thus further improve the localization performance. In this paper, we propose a novel reconstruction-based method named knowledge-informed self-training (KIST) which integrates knowledge into reconstruction model through self-training. Specifically, KIST utilizes weakly labeled anomalous samples in addition to the normal ones and exploits knowledge to yield pixel-level pseudo-labels of the anomalous samples. Based on the pseudo labels, a novel loss which promotes the reconstruction of normal pixels while suppressing the reconstruction of anomalous pixels is used. We conduct experiments on different datasets and demonstrate the advantages of KIST over the existing reconstruction-based methods.

Reconstruction-Based Anomaly Localization via Knowledge-Informed Self-Training

TL;DR

A novel reconstruction-based method named knowledge-informed self-training (KIST) which integrates knowledge into a reconstruction model through self-training and exploits knowledge to yield pixel-level pseudolabels of the anomalous samples.

Abstract

Anomaly localization, which involves localizing anomalous regions within images, is a significant industrial task. Reconstruction-based methods are widely adopted for anomaly localization because of their low complexity and high interpretability. Most existing reconstruction-based methods only use normal samples to construct model. If anomalous samples are appropriately utilized in the process of anomaly localization, the localization performance can be improved. However, usually only weakly labeled anomalous samples are available, which limits the improvement. In many cases, we can obtain some knowledge of anomalies summarized by domain experts. Taking advantage of such knowledge can help us better utilize the anomalous samples and thus further improve the localization performance. In this paper, we propose a novel reconstruction-based method named knowledge-informed self-training (KIST) which integrates knowledge into reconstruction model through self-training. Specifically, KIST utilizes weakly labeled anomalous samples in addition to the normal ones and exploits knowledge to yield pixel-level pseudo-labels of the anomalous samples. Based on the pseudo labels, a novel loss which promotes the reconstruction of normal pixels while suppressing the reconstruction of anomalous pixels is used. We conduct experiments on different datasets and demonstrate the advantages of KIST over the existing reconstruction-based methods.
Paper Structure (23 sections, 31 equations, 8 figures, 3 tables, 1 algorithm)

This paper contains 23 sections, 31 equations, 8 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. KIST Method
  3. Overview of KIST
  4. Knowledge-Based Anomaly Grade Evaluation
  5. Knowledge representation using fuzzy rule
  6. Anomaly grade evaluation
  7. Knowledge-Informed Self-Training
  8. Reconstruction model based on neural network
  9. Initial model constructing using normal samples
  10. Pseudo-label producing for anomalous samples
  11. Model updating by contrastive reconstruction
  12. Overall Algorithm of KIST
  13. Guided Filter-Based Post Processing
  14. Experiments
  15. Experimental Setup
  16. ...and 8 more sections

Figures (8)

  • Figure 1: Flowchart of KIST-based anomaly localization scheme.
  • Figure 2: Diagram of KIST.
  • Figure 3: Examples of anomalous samples (the anomalous regions are contoured with red line). The anomalous region of (a) has low gray value and large area. The anomalous region of (b) has high gray value and rectangle shape. The anomalous region of (c) has low gray value and slender shape. The anomalous region of (d) has low gray value and small area.
  • Figure 4: Diagram of pseudo-label producing process.
  • Figure 5: Trapezoidal membership functions (trapmfs) used for calculating the membership grades of the property value.
  • ...and 3 more figures

Theorems & Definitions (2)

  • Remark 1
  • Remark 2