PANDA: Adapting Pretrained Features for Anomaly Detection and Segmentation

TL;DR

PANDA demonstrates that adapting pretrained features with simple methods significantly improves anomaly detection and segmentation, outperforming complex state-of-the-art techniques across various settings.

Contribution

The paper introduces PANDA, a novel approach that adapts pretrained features for anomaly detection, addressing feature collapse with new regularization and stopping strategies.

Findings

Pretrained features combined with simple methods outperform complex models.

PANDA outperforms state-of-the-art in anomaly detection and segmentation.

Proposed adaptation methods prevent feature collapse effectively.

Abstract

Anomaly detection methods require high-quality features. In recent years, the anomaly detection community has attempted to obtain better features using advances in deep self-supervised feature learning. Surprisingly, a very promising direction, using pretrained deep features, has been mostly overlooked. In this paper, we first empirically establish the perhaps expected, but unreported result, that combining pretrained features with simple anomaly detection and segmentation methods convincingly outperforms, much more complex, state-of-the-art methods. In order to obtain further performance gains in anomaly detection, we adapt pretrained features to the target distribution. Although transfer learning methods are well established in multi-class classification problems, the one-class classification (OCC) setting is not as well explored. It turns out that naive adaptation methods, which typically work well in supervised learning, often result in catastrophic collapse (feature deterioration) and reduce performance in OCC settings. A popular OCC method, DeepSVDD, advocates using specialized architectures, but this limits the adaptation performance gain. We propose two methods for combating collapse: i) a variant of early stopping that dynamically learns the stopping iteration ii) elastic regularization inspired by continual learning. Our method, PANDA, outperforms the state-of-the-art in the OCC, outlier exposure and anomaly segmentation settings by large margins.