Anomaly Detection via Reverse Distillation from One-Class Embedding
This addresses the problem of improving anomaly detection accuracy for applications like industrial inspection or medical diagnosis, though it appears to be an incremental improvement over existing knowledge distillation approaches.
The paper tackles the problem of limited anomalous representation diversity in knowledge distillation-based anomaly detection by proposing a reverse distillation paradigm where a student decoder reconstructs a teacher encoder's multiscale representations from a one-class embedding. The method achieves state-of-the-art performance on anomaly detection and one-class novelty detection benchmarks.
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.