Article
Version 1
Preserved in Portico This version is not peer-reviewed
Action Recognition via Adaptive Semi-Supervised Feature Analysis
Version 1
: Received: 26 May 2023 / Approved: 29 May 2023 / Online: 29 May 2023 (03:34:35 CEST)
A peer-reviewed article of this Preprint also exists.
Xu, Z.; Li, X.; Li, J.; Chen, H.; Hu, R. Action Recognition via Adaptive Semi-Supervised Feature Analysis. Appl. Sci. 2023, 13, 7684. Xu, Z.; Li, X.; Li, J.; Chen, H.; Hu, R. Action Recognition via Adaptive Semi-Supervised Feature Analysis. Appl. Sci. 2023, 13, 7684.
Abstract
This paper presents a new algorithm for semi-supervised action recognition using adaptive feature analysis. We model the samples of the same action as the same manifold and those of different actions as different manifolds, and integrate feature correlation mining and visual similarity mining into a joint framework. By maximizing the separability between different classes based on the labeled data, the proposed algorithm learns multiple features from labeled data and utilizes the intrinsic geometric structure of the data distribution between labeled and unlabeled data to boost the recognition. We introduce Projected Barzilai-Borwein (PBB) to train a subspace projection matrix as classifier. Extensive experiments on three benchmark datasets demonstrate that our algorithm outperforms the existing semi-supervised action recognition algorithms when there are a few labeled training samples.
Keywords
Nonmonotone line search; Two-point stepsize gradient; Grassmannian kernels
Subject
Computer Science and Mathematics, Computer Vision and Graphics
Copyright: This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Comments (0)
We encourage comments and feedback from a broad range of readers. See criteria for comments and our Diversity statement.
Leave a public commentSend a private comment to the author(s)
* All users must log in before leaving a comment