Rethinking Assumptions in Deep Anomaly Detection
This work addresses anomaly detection for image data, showing that minimal labeled anomalies can significantly improve performance, which is incremental but impactful for the field.
The paper tackled the problem of deep anomaly detection on images by challenging the assumption that unsupervised methods are necessary due to lack of anomalous data, and found that classifiers trained with only 64 random natural images as anomalies outperformed the state of the art on an ImageNet benchmark.
Though anomaly detection (AD) can be viewed as a classification problem (nominal vs. anomalous) it is usually treated in an unsupervised manner since one typically does not have access to, or it is infeasible to utilize, a dataset that sufficiently characterizes what it means to be "anomalous." In this paper we present results demonstrating that this intuition surprisingly seems not to extend to deep AD on images. For a recent AD benchmark on ImageNet, classifiers trained to discern between normal samples and just a few (64) random natural images are able to outperform the current state of the art in deep AD. Experimentally we discover that the multiscale structure of image data makes example anomalies exceptionally informative.