Diversified Fisher kernel: encoding discrimination in Fisher features to compete deep neural models for visual classification task. Issue 8 (23rd November 2020)

• Record Type: Journal Article
• Authors: Ahmed, Sarah; Azim, Tayyaba
• Is Part Of: IET computer vision. Volume 14:Issue 8(2020)
• Page Start: 658
• Page End: 664
• Publication Date: 2020-11-23
• Subjects: neural nets -- pattern classification -- Boltzmann machines -- learning (artificial intelligence) -- feature extraction -- image representation -- image classification; large‐scale visual retrieval -- feature selection technique -- discriminative feature representations -- dimensional Fisher features -- kernel learning -- classification performance -- state‐of‐the‐art feature selection techniques -- reduced Fisher space -- diversified Fisher kernel -- deep neural models -- visual classification task -- stochastic probabilistic models -- restricted Boltzmann machines -- deep Boltzmann machines -- competitive visual classification results -- widely popular deep discriminative models -- Fisher kernels bridges -- deep learning paradigm -- deep architecture -- discriminative classifiers -- Fisher vectors; Computer vision -- Periodicals; Pattern recognition systems -- Periodicals; 006.37
• DOI: 10.1049/iet-cvi.2019.0208
• Languages: English
• ISSNs: 1751-9632

Abstract

Abstract : Fisher kernels derived from stochastic probabilistic models such as restricted and deep Boltzmann machines have shown competitive visual classification results in comparison to widely popular deep discriminative models. This genre of Fisher kernels bridges the gap between shallow and deep learning paradigm by inducing the characteristics of deep architecture into Fisher kernel, further deployed for classification in discriminative classifiers. Despite their success, the memory and computational costs of Fisher vectors do not make them amenable for large‐scale visual retrieval and classification tasks. This study introduces a novel feature selection technique inspired from the functional characteristics of neural architectures for learning discriminative feature representations to boost the performance of Fisher kernels against deep discriminative models. The proposed technique condenses the large dimensional Fisher features for kernel learning and shows improvement in its classification performance and storage cost on leading benchmark data sets. A comparison of the proposed method with other state‐of‐the‐art feature selection techniques is made to demonstrate its performance supremacy as well as time complexity required to learn in reduced Fisher space.