Semi-supervised Fisher vector network

TL;DR

This paper introduces a semi-supervised Fisher vector network that combines unsupervised and supervised learning to improve image classification and action recognition by leveraging unlabeled data.

Contribution

It proposes a hybrid neural network architecture that models Fisher vector pipeline components, enabling semi-supervised learning for better feature representation.

Findings

Performance improves with more unlabeled data

Effective for image classification tasks

Enhances action recognition accuracy

Abstract

In this work we explore how the architecture proposed in [8], which expresses the processing steps of the classical Fisher vector pipeline approaches, i.e. dimensionality reduction by principal component analysis (PCA) projection, Gaussian mixture model (GMM) and Fisher vector descriptor extraction as network layers, can be modified into a hybrid network that combines the benefits of both unsupervised and supervised training methods, resulting in a model that learns a semi-supervised Fisher vector descriptor of the input data. We evaluate the proposed model at image classification and action recognition problems and show how the model's classification performance improves as the amount of unlabeled data increases during training.