Show simple item record

ASNet: Auto-Augmented Siamese Neural Network for Action Recognition

dc.contributor.authorYujia Zhang
dc.contributor.authorLai-Man Po
dc.contributor.authorJingjing Xiong
dc.contributor.authorYasar Abbas Ur REHMAN
dc.contributor.authorKwok-Wai Cheung
dc.contributor.otherDepartment of Electrical Engineering, City University of Hong Kong, Hong Kong, China
dc.contributor.otherDepartment of Electrical Engineering, City University of Hong Kong, Hong Kong, China
dc.contributor.otherDepartment of Electrical Engineering, City University of Hong Kong, Hong Kong, China
dc.contributor.otherTCL Corporate Research Co. Limited, Hong Kong, China
dc.contributor.otherSchool of Communication, The Hang Seng University of Hong Kong, Hong Kong, China
dc.date.accessioned2025-10-09T04:58:48Z
dc.date.available2025-10-09T04:58:48Z
dc.date.issued01-07-2021
dc.identifier.urihttps://www.mdpi.com/1424-8220/21/14/4720
dc.identifier.urihttp://digilib.fisipol.ugm.ac.id/repo/handle/15717717/40860
dc.description.abstractHuman action recognition methods in videos based on deep convolutional neural networks usually use random cropping or its variants for data augmentation. However, this traditional data augmentation approach may generate many non-informative samples (video patches covering only a small part of the foreground or only the background) that are not related to a specific action. These samples can be regarded as noisy samples with incorrect labels, which reduces the overall action recognition performance. In this paper, we attempt to mitigate the impact of noisy samples by proposing an Auto-augmented Siamese Neural Network (ASNet). In this framework, we propose backpropagating salient patches and randomly cropped samples in the same iteration to perform gradient compensation to alleviate the adverse gradient effects of non-informative samples. Salient patches refer to the samples containing critical information for human action recognition. The generation of salient patches is formulated as a Markov decision process, and a reinforcement learning agent called SPA (Salient Patch Agent) is introduced to extract patches in a weakly supervised manner without extra labels. Extensive experiments were conducted on two well-known datasets UCF-101 and HMDB-51 to verify the effectiveness of the proposed SPA and ASNet.
dc.language.isoEN
dc.publisherMDPI AG
dc.subject.lccChemical technology
dc.titleASNet: Auto-Augmented Siamese Neural Network for Action Recognition
dc.typeArticle
dc.description.keywordsaction recognition
dc.description.keywords3D-CNN
dc.description.keywordsdeep reinforcement learning
dc.description.keywordsdata augmentation
dc.description.doi10.3390/s21144720
dc.title.journalSensors
dc.identifier.e-issn1424-8220
dc.identifier.oaioai:doaj.org/journal:b25f04f1d387428caba56ce4a563a710
dc.journal.infoVolume 21, Issue 14


This item appears in the following Collection(s)

Show simple item record