Iterative energy-based projection on a normal data manifold for anomaly localization

TL;DR

This paper introduces an iterative energy-based projection method on a normal data manifold to improve anomaly localization and image inpainting, surpassing traditional autoencoder reconstructions in quality.

Contribution

The paper presents a novel iterative projection technique using gradient descent on an energy function, enhancing anomaly localization and image quality beyond standard autoencoder methods.

Findings

Achieves state-of-the-art results on anomaly localization datasets.

Produces higher quality images than classic autoencoder reconstructions.

Shows promising inpainting results on CelebA dataset.

Abstract

Autoencoder reconstructions are widely used for the task of unsupervised anomaly localization. Indeed, an autoencoder trained on normal data is expected to only be able to reconstruct normal features of the data, allowing the segmentation of anomalous pixels in an image via a simple comparison between the image and its autoencoder reconstruction. In practice however, local defects added to a normal image can deteriorate the whole reconstruction, making this segmentation challenging. To tackle the issue, we propose in this paper a new approach for projecting anomalous data on a autoencoder-learned normal data manifold, by using gradient descent on an energy derived from the autoencoder's loss function. This energy can be augmented with regularization terms that model priors on what constitutes the user-defined optimal projection. By iteratively updating the input of the autoencoder, we bypass the loss of high-frequency information caused by the autoencoder bottleneck. This allows to produce images of higher quality than classic reconstructions. Our method achieves state-of-the-art results on various anomaly localization datasets. It also shows promising results at an inpainting task on the CelebA dataset.