Two Is Better Than One: Aligned Representation Pairs for Anomaly Detection

TL;DR

This paper introduces Con$_2$, a novel self-supervised approach that leverages symmetry-based context and content alignment to learn representations for anomaly detection, especially effective in specialized medical datasets.

Contribution

The paper proposes Con$_2$, a new method using symmetry-based context contrast and content alignment for improved anomaly detection without prior knowledge of unseen data.

Findings

Outperforms existing self-supervised and pretrained methods on medical datasets.

Achieves competitive results on natural imaging benchmarks.

Effectively detects anomalies as outliers in learned context clusters.

Abstract

Anomaly detection focuses on identifying samples that deviate from the norm. Discovering informative representations of normal samples is crucial to detecting anomalies effectively. Recent self-supervised methods have successfully learned such representations by employing prior knowledge about anomalies to create synthetic outliers during training. However, we often do not know what to expect from unseen data in specialized real-world applications. In this work, we address this limitation with our new approach Con$_2$, which leverages prior knowledge about symmetries in normal samples to observe the data in different contexts. Con$_2$ consists of two parts: Context Contrasting clusters representations according to their context, while Content Alignment encourages the model to capture semantic information by aligning the positions of normal samples across clusters. The resulting representation space allows us to detect anomalies as outliers of the learned context clusters. We demonstrate the benefit of this approach in extensive experiments on specialized medical datasets, outperforming competitive baselines based on self-supervised learning and pretrained models and presenting competitive performance on natural imaging benchmarks.