Improving Unsupervised Defect Segmentation by Applying Structural Similarity to Autoencoders

Abstract

Convolutional autoencoders have emerged as popular methods for unsupervised defect segmentation on image data. Most commonly, this task is performed by thresholding a pixel-wise reconstruction error based on an ${\ell }^p$ distance. This procedure, however, leads to large residuals whenever the reconstruction encompasses slight localization inaccuracies around edges. It also fails to reveal defective regions that have been visually altered when intensity values stay roughly consistent. We show that these problems prevent these approaches from being applied to complex real-world scenarios and that it cannot be easily avoided by employing more elaborate architectures such as variational or feature matching autoencoders. We propose to use a perceptual loss function based on structural similarity which examines inter-dependencies between local image regions, taking into account luminance, contrast and structural information, instead of simply comparing single pixel values. It achieves significant performance gains on a challenging real-world dataset of nanofibrous materials and a novel dataset of two woven fabrics over the state of the art approaches for unsupervised defect segmentation that use pixel-wise reconstruction error metrics.

References in corpus (5)

• Adam: A Method for Stochastic Optimization
• Auto-Encoding Variational Bayes
• Efficient GAN-Based Anomaly Detection
• Deep Autoencoding Models for Unsupervised Anomaly Segmentation in Brain MR Images
• Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery

Cited by in corpus (47)

• Uninformed Students: Student-Teacher Anomaly Detection with Discriminative Latent Embeddings
• Deep Industrial Image Anomaly Detection: A Survey
• Omni-frequency Channel-selection Representations for Unsupervised Anomaly Detection
• Deep Learning for Unsupervised Anomaly Localization in Industrial Images: A Survey
• DFR: Deep Feature Reconstruction for Unsupervised Anomaly Segmentation
• Advances in Deep Concealed Scene Understanding
• A Survey on Unsupervised Anomaly Detection Algorithms for Industrial Images
• Student-Teacher Feature Pyramid Matching for Anomaly Detection
• Segmentation of cell-level anomalies in electroluminescence images of photovoltaic modules
• MOCCA: Multi-Layer One-Class ClassificAtion for Anomaly Detection
• InsPLAD: A Dataset and Benchmark for Power Line Asset Inspection in UAV Images
• Patch SVDD: Patch-level SVDD for Anomaly Detection and Segmentation
• DRAEM -- A discriminatively trained reconstruction embedding for surface anomaly detection
• Masked Transformer for image Anomaly Localization
• Unsupervised Pathology Detection: A Deep Dive Into the State of the Art
• Anomaly localization by modeling perceptual features
• Challenging Current Semi-Supervised Anomaly Segmentation Methods for Brain MRI
• Iterative energy-based projection on a normal data manifold for anomaly localization
• The MVTec AD 2 Dataset: Advanced Scenarios for Unsupervised Anomaly Detection
• Unsupervised Two-Stage Anomaly Detection
• Unsupervised Abnormality Detection through Mixed Structure Regularization (MSR) in Deep Sparse Autoencoders
• VT-ADL: A Vision Transformer Network for Image Anomaly Detection and Localization
• AutoOD: Automated Outlier Detection via Curiosity-guided Search and Self-imitation Learning
• Unsupervised Anomaly Localization with Structural Feature-Autoencoders
• Puzzle-AE: Novelty Detection in Images through Solving Puzzles
• Contrastive Predictive Coding for Anomaly Detection
• Rethinking Polyp Segmentation from an Out-of-Distribution Perspective
• A Continual Learning Framework for Adaptive Defect Classification and Inspection
• Towards Phytoplankton Parasite Detection Using Autoencoders
• Innovations Autoencoder and its Application in One-class Anomalous Sequence Detection
• Label-Free Segmentation of COVID-19 Lesions in Lung CT
• MLF-SC: Incorporating multi-layer features to sparse coding for anomaly detection
• TrustMAE: A Noise-Resilient Defect Classification Framework using Memory-Augmented Auto-Encoders with Trust Regions
• Improving unsupervised anomaly localization by applying multi-scale memories to autoencoders
• GRD-Net: Generative-Reconstructive-Discriminative Anomaly Detection with Region of Interest Attention Module
• Learning Global and Local Features of Normal Brain Anatomy for Unsupervised Abnormality Detection
• Image-based Plant Disease Diagnosis with Unsupervised Anomaly Detection Based on Reconstructability of Colors
• Anomaly Detection By Autoencoder Based On Weighted Frequency Domain Loss
• ISP-AD: A Large-Scale Real-World Dataset for Advancing Industrial Anomaly Detection with Synthetic and Real Defects
• AdeNet: Deep learning architecture that identifies damaged electrical insulators in power lines
• Energy-Based Anomaly Detection and Localization
• Autoencoding Under Normalization Constraints
• A Survey on Proactive Customer Care: Enabling Science and Steps to Realize it
• Teacher Encoder-Student Decoder Denoising Guided Segmentation Network for Anomaly Detection
• Data augmentation and pre-trained networks for extremely low data regimes unsupervised visual inspection
• P-WAE: Generalized Patch-Wasserstein Autoencoder for Anomaly Screening
• Computer Vision and Normalizing Flow-Based Defect Detection