Unified Anomaly Detection via Multi-Scale Contrasted Memory

Loïc Jézéquel, Ngoc-Son Vu, Jean Beaudet, Aymeric Histace

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Keywords: anomaly detection, self-supervised learning, unbalanced outlier-exposure, hopfield memory

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

Abstract:

Deep anomaly detection (AD) aims to provide robust and efficient classifiers for one-class (OC) and outlier-exposure (OE) settings. However current models still struggle on edge-case normal samples and are often unable to keep high performance over different scales of anomalies. Additionally, there is a lack of a unified framework that efficiently addresses both OC and OE settings. To address these limitations, we present a novel two-stage method which leverages multi-scale normal prototypes during training to compute an anomaly deviation score. First, we employ a novel memory-augmented contrastive learning (CL) to jointly learn representations and memory modules across multiple scales. This allows us to effectively capture subtle features of normal data while adapting to varying levels of anomaly complexity.
Then, we train an efficient anomaly distance-based detector that computes spatial deviation maps between the learned prototypes and incoming observations.
Our model outperforms the state-of-the-art on a wide range of anomalies, including object, style, and local anomalies, as well as face presentation attacks. Notably, it stands as the first model capable of maintaining exceptional performance across both OC and OE settings.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 5400