Uninformed Students: Student-Teacher Anomaly Detection with Discriminative Latent Embeddings

Abstract

We introduce a powerful student-teacher framework for the challenging problem of unsupervised anomaly detection and pixel-precise anomaly segmentation in high-resolution images. Student networks are trained to regress the output of a descriptive teacher network that was pretrained on a large dataset of patches from natural images. This circumvents the need for prior data annotation. Anomalies are detected when the outputs of the student networks differ from that of the teacher network. This happens when they fail to generalize outside the manifold of anomaly-free training data. The intrinsic uncertainty in the student networks is used as an additional scoring function that indicates anomalies. We compare our method to a large number of existing deep learning based methods for unsupervised anomaly detection. Our experiments demonstrate improvements over state-of-the-art methods on a number of real-world datasets, including the recently introduced MVTec Anomaly Detection dataset that was specifically designed to benchmark anomaly segmentation algorithms.

Accepted to CVPR 2020

References in corpus (8)

• Adam: A Method for Stochastic Optimization
• What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?
• Improving Unsupervised Defect Segmentation by Applying Structural Similarity to Autoencoders
• Deep Autoencoding Models for Unsupervised Anomaly Segmentation in Brain MR Images
• Anomaly Detection using One-Class Neural Networks
• Exploiting Epistemic Uncertainty of Anatomy Segmentation for Anomaly Detection in Retinal OCT
• Metric Learning for Novelty and Anomaly Detection
• Are pre-trained CNNs good feature extractors for anomaly detection in surveillance videos?

Cited by in corpus (67)

• Knowledge Distillation: A Survey
• Deep Industrial Image Anomaly Detection: A Survey
• Mixed supervision for surface-defect detection: from weakly to fully supervised learning
• Sub-Image Anomaly Detection with Deep Pyramid Correspondences
• 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
• Collaborative Discrepancy Optimization for Reliable Image Anomaly Localization
• A Survey on Unsupervised Anomaly Detection Algorithms for Industrial Images
• Self-Supervised Masked Convolutional Transformer Block for Anomaly Detection
• A Self-Reasoning Framework for Anomaly Detection Using Video-Level Labels
• Student-Teacher Feature Pyramid Matching for Anomaly Detection
• One-Class Classification: A Survey
• MOCCA: Multi-Layer One-Class ClassificAtion for Anomaly Detection
• Deep Graph-level Anomaly Detection by Glocal Knowledge Distillation
• Dual-distribution discrepancy with self-supervised refinement for anomaly detection in medical images
• Attribute Restoration Framework for Anomaly Detection
• Explainable Deep Few-shot Anomaly Detection with Deviation Networks
• AnoSeg: Anomaly Segmentation Network Using Self-Supervised Learning
• DRAEM -- A discriminatively trained reconstruction embedding for surface anomaly detection
• Masked Transformer for image Anomaly Localization
• Progressive Boundary Guided Anomaly Synthesis for Industrial Anomaly Detection
• Unsupervised Pathology Detection: A Deep Dive Into the State of the Art
• Anomaly Detection with Conditioned Denoising Diffusion Models
• Anomaly localization by modeling perceptual features
• Learning image representations for anomaly detection: application to discovery of histological alterations in drug development
• SoftPatch+: Fully Unsupervised Anomaly Classification and Segmentation
• The MVTec AD 2 Dataset: Advanced Scenarios for Unsupervised Anomaly Detection
• PKU-GoodsAD: A Supermarket Goods Dataset for Unsupervised Anomaly Detection and Segmentation
• Dual-Modeling Decouple Distillation for Unsupervised Anomaly Detection
• FAIR: Frequency-aware Image Restoration for Industrial Visual Anomaly Detection
• Reconstruction from edge image combined with color and gradient difference for industrial surface anomaly detection
• Towards High-resolution 3D Anomaly Detection via Group-Level Feature Contrastive Learning
• A Unified Survey on Anomaly, Novelty, Open-Set, and Out-of-Distribution Detection: Solutions and Future Challenges
• Unsupervised Two-Stage Anomaly Detection
• Reconstruction Student with Attention for Student-Teacher Pyramid Matching
• PUAD: Frustratingly Simple Method for Robust Anomaly Detection
• Visual Anomaly Detection for Images: A Survey
• Unsupervised Anomaly Detection with an Enhanced Teacher for Student-Teacher Feature Pyramid Matching
• VT-ADL: A Vision Transformer Network for Image Anomaly Detection and Localization
• Transfer Learning Gaussian Anomaly Detection by Fine-tuning Representations
• Dual-path Frequency Discriminators for Few-shot Anomaly Detection
• Explainable AI guided unsupervised fault diagnostics for high-voltage circuit breakers
• Contrastive Predictive Coding for Anomaly Detection
• Deep Autoencoders for Anomaly Detection in Textured Images using CW-SSIM
• A Prototype-Based Neural Network for Image Anomaly Detection and Localization
• Identifying the Defective: Detecting Damaged Grains for Cereal Appearance Inspection
• Lightning Fast Video Anomaly Detection via Adversarial Knowledge Distillation
• CVTGAD: Simplified Transformer with Cross-View Attention for Unsupervised Graph-level Anomaly Detection
• Improving Novelty Detection using the Reconstructions of Nearest Neighbours
• Focus Your Distribution: Coarse-to-Fine Non-Contrastive Learning for Anomaly Detection and Localization
• SQUID: Deep Feature In-Painting for Unsupervised Anomaly Detection
• Label-Free Segmentation of COVID-19 Lesions in Lung CT
• SLA${}^2$P: Self-supervised Anomaly Detection with Adversarial Perturbation
• GRD-Net: Generative-Reconstructive-Discriminative Anomaly Detection with Region of Interest Attention Module
• Improving unsupervised anomaly localization by applying multi-scale memories to autoencoders
• Anomaly Detection via Self-organizing Map
• Attention Map-guided Two-stage Anomaly Detection using Hard Augmentation
• Anomaly Detection-Based UE-Centric Inter-Cell Interference Suppression
• ISP-AD: A Large-Scale Real-World Dataset for Advancing Industrial Anomaly Detection with Synthetic and Real Defects
• Critical Review for One-class Classification: recent advances and the reality behind them
• Teacher Encoder-Student Decoder Denoising Guided Segmentation Network for Anomaly Detection
• Evaluating the Use of Reconstruction Error for Novelty Localization
• Brittle Features May Help Anomaly Detection
• Learning to Be a Transformer to Pinpoint Anomalies
• Data augmentation and pre-trained networks for extremely low data regimes unsupervised visual inspection
• Y-GAN: Learning Dual Data Representations for Efficient Anomaly Detection