Anomaly Detection via Reverse Distillation from One-Class Embedding

TL;DR

This paper introduces a reverse distillation approach using a teacher encoder and student decoder with a one-class embedding, significantly improving unsupervised anomaly detection performance.

Contribution

It proposes a novel reverse distillation paradigm and a trainable one-class bottleneck embedding to enhance anomaly detection accuracy and generalizability.

Findings

Outperforms state-of-the-art on AD benchmarks

Effectively preserves normal patterns while ignoring anomalies

Demonstrates strong generalizability across datasets

Abstract

Knowledge distillation (KD) achieves promising results on the challenging problem of unsupervised anomaly detection (AD).The representation discrepancy of anomalies in the teacher-student (T-S) model provides essential evidence for AD. However, using similar or identical architectures to build the teacher and student models in previous studies hinders the diversity of anomalous representations. To tackle this problem, we propose a novel T-S model consisting of a teacher encoder and a student decoder and introduce a simple yet effective "reverse distillation" paradigm accordingly. Instead of receiving raw images directly, the student network takes teacher model's one-class embedding as input and targets to restore the teacher's multiscale representations. Inherently, knowledge distillation in this study starts from abstract, high-level presentations to low-level features. In addition, we introduce a trainable one-class bottleneck embedding (OCBE) module in our T-S model. The obtained compact embedding effectively preserves essential information on normal patterns, but abandons anomaly perturbations. Extensive experimentation on AD and one-class novelty detection benchmarks shows that our method surpasses SOTA performance, demonstrating our proposed approach's effectiveness and generalizability.