Open-Set Recognition with Gaussian Mixture Variational Autoencoders

TL;DR

This paper introduces a Gaussian mixture variational autoencoder that improves open-set classification by jointly learning reconstruction and class-based clustering, leading to more accurate detection of unknown classes.

Contribution

It proposes a novel GMVAE model that enhances open-set recognition by integrating reconstruction and clustering in the latent space, unlike previous methods.

Findings

Achieves 29.5% higher average F1 score in open-set classification

Demonstrates robustness across extensive experiments

Provides analytical insights into model performance

Abstract

In inference, open-set classification is to either classify a sample into a known class from training or reject it as an unknown class. Existing deep open-set classifiers train explicit closed-set classifiers, in some cases disjointly utilizing reconstruction, which we find dilutes the latent representation's ability to distinguish unknown classes. In contrast, we train our model to cooperatively learn reconstruction and perform class-based clustering in the latent space. With this, our Gaussian mixture variational autoencoder (GMVAE) achieves more accurate and robust open-set classification results, with an average F1 improvement of 29.5%, through extensive experiments aided by analytical results.