Conditional Gaussian Distribution Learning for Open Set Recognition

TL;DR

This paper introduces Conditional Gaussian Distribution Learning (CGDL), a novel approach for open set recognition that improves unknown detection and known class classification by modeling latent features with Gaussian distributions.

Contribution

The paper proposes CGDL, combining Gaussian modeling of latent features with a probabilistic ladder architecture to enhance open set recognition performance.

Findings

Outperforms baseline methods on standard image datasets

Achieves state-of-the-art results in open set recognition

Effectively detects unknown samples while classifying known ones

Abstract

Deep neural networks have achieved state-of-the-art performance in a wide range of recognition/classification tasks. However, when applying deep learning to real-world applications, there are still multiple challenges. A typical challenge is that unknown samples may be fed into the system during the testing phase and traditional deep neural networks will wrongly recognize the unknown sample as one of the known classes. Open set recognition is a potential solution to overcome this problem, where the open set classifier should have the ability to reject unknown samples as well as maintain high classification accuracy on known classes. The variational auto-encoder (VAE) is a popular model to detect unknowns, but it cannot provide discriminative representations for known classification. In this paper, we propose a novel method, Conditional Gaussian Distribution Learning (CGDL), for open set recognition. In addition to detecting unknown samples, this method can also classify known samples by forcing different latent features to approximate different Gaussian models. Meanwhile, to avoid information hidden in the input vanishing in the middle layers, we also adopt the probabilistic ladder architecture to extract high-level abstract features. Experiments on several standard image datasets reveal that the proposed method significantly outperforms the baseline method and achieves new state-of-the-art results.