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

By Paul Bergmann, Michael Fauser, David Sattlegger, Carsten Steger. 2019

971 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Knowledge Distillation: A Survey (2020, background): “, 2020c), anomaly detection (Bergmann et al., 2020), facial landmark detection (Dong and Yang, 2019), video classification (Bhardwaj et al.”
DeSTSeg: Segmentation Guided Denoising Student-Teacher for Anomaly Detection (2022, background): “For example, [3] used ensemble learning to train multiple student networks and exploited the irregularity of their feature representations to recognize the anomaly.”
SPot-the-Difference Self-Supervised Pre-training for Anomaly Detection and Segmentation (2022, methodology): “In [3], the best method is Student-Teacher [5] with 92.2% AU-ROC which seems to be close to perfection.”
CFLOW-AD: Real-Time Unsupervised Anomaly Detection with Localization via Conditional Normalizing Flows (2021, background): “Another line of research proposes to employ a studentteacher type of framework [5, 33, 41].”
Deep Industrial Image Anomaly Detection: A Survey (2023, neutral): “Bergmann et al. [9] is the first to use teacher-student architecture for anomaly detection.”
PANDA: Adapting Pretrained Features for Anomaly Detection and Segmentation (2020, methodology): “[3] used a student-teacher based autoencoder approach employing pre-trained ImageNet deep features.”
VT-ADL: A Vision Transformer Network for Image Anomaly Detection and Localization (2021, methodology): “Then we employed the PRO (Per Region Overlap) [8], [24] as the evaluation metric for the MVTec and BTAD datasets.”
Sub-Image Anomaly Detection with Deep Pyramid Correspondences (2020, methodology): “Our method is compared with the auto-encoder with pre-trained features based approach of Bregmann et al. [6] and the baselines presented in their paper.”
DRÆM – A discriminatively trained reconstruction embedding for surface anomaly detection (2021, methodology): “Class US[4] RIAD[31]”
CFA: Coupled-Hypersphere-Based Feature Adaptation for Target-Oriented Anomaly Localization (2022, methodology): “For example, CFA++ showed 10.1% higher P-AUPRO score than PaDiM.”
PaDiM: a Patch Distribution Modeling Framework for Anomaly Detection and Localization (2020, methodology): “Type Reconstruction-based methods Embedding similarity based methods Our methods Model AE simm [1], [2], [9] AE L2 [1], [2] VAE Student [2] Patch SVDD [4] SPADE [5] PaDiMR18-Rd100 PaDiMWR50-Rd550 Carpet (87, 64.”
Towards Total Recall in Industrial Anomaly Detection (2021, methodology): “The PRO score takes into account the overlap and recovery of connected anomaly components to better account for varying anomaly sizes in MVTec AD, see [6] for details.”
AnomalyCLIP: Object-agnostic Prompt Learning for Zero-shot Anomaly Detection (2023, neutral): “…industrial defect inspection (Bergmann et al., 2019; Xie et al., 2023; Roth et al., 2022; Huang et al., 2022; Mou et al., 2022; Chen et al., 2022; Bergmann et al., 2020; Pang et al., 2021a; Reiss & Hoshen, 2023; You et al., 2022; Liznerski et al., 2020; Ding et al., 2022; Zhou et al., 2023; Cao…”
EfficientAD: Accurate Visual Anomaly Detection at Millisecond-Level Latencies (2023, methodology): “…student–teacher pair, however, lacks techniques used by previous methods to increase the anomaly detection performance: ensembling multiple teachers and students [10], using features from a pyramid of layers [62], and using architectural asymmetry between the student and the teacher network [50].”
A Unified Model for Multi-class Anomaly Detection (2022, background): “US [3] PSVDD [15] PaDiM [5] CutPaste [10] FCDD [11] MKD [12] DRAEM [16] Ours FLOPs(G) 60.”
Multiresolution Knowledge Distillation for Anomaly Detection (2020, background): “[8] train an ensemble of student networks to mimic the last layer of a teacher network on the anomaly-free data.”