Frame Aggregation and Multi-Modal Fusion Framework for Video-Based Person Recognition

Video-based person recognition is challenging due to persons being blocked and blurred, and the variation of shooting angle. Previous research always focused on person recognition on still images, ignoring similarity and continuity between video frames. To tackle the challenges above, we propose a n...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Li, Fangtao, Wang, Wenzhe, Liu, Zihe, Wang, Haoran, Yan, Chenghao, Wu, Bin
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Li, Fangtao
Wang, Wenzhe
Liu, Zihe
Wang, Haoran
Yan, Chenghao
Wu, Bin
description Video-based person recognition is challenging due to persons being blocked and blurred, and the variation of shooting angle. Previous research always focused on person recognition on still images, ignoring similarity and continuity between video frames. To tackle the challenges above, we propose a novel Frame Aggregation and Multi-Modal Fusion (FAMF) framework for video-based person recognition, which aggregates face features and incorporates them with multi-modal information to identify persons in videos. For frame aggregation, we propose a novel trainable layer based on NetVLAD (named AttentionVLAD), which takes arbitrary number of features as input and computes a fixed-length aggregation feature based on feature quality. We show that introducing an attention mechanism to NetVLAD can effectively decrease the impact of low-quality frames. For the multi-model information of videos, we propose a Multi-Layer Multi-Modal Attention (MLMA) module to learn the correlation of multi-modality by adaptively updating Gram matrix. Experimental results on iQIYI-VID-2019 dataset show that our framework outperforms other state-of-the-art methods.
doi_str_mv 10.48550/arxiv.2010.09290
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2010_09290</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2010_09290</sourcerecordid><originalsourceid>FETCH-LOGICAL-a670-9ce6b7390a88d42cd8523111369e6c54ed5303a4a3e1835c6c3f1eb400ba76073</originalsourceid><addsrcrecordid>eNotj8tOwzAQRb1hgQofwAr_gMs4jh1nWSoCSK1AqLCNJvYkskhj5LQ8_p4msLrS0dWRDmNXEpa51RpuMH2Hz2UGJwBlVsI5e60S7omvui5Rh4cQB46D59tjfwhiGz32vDqOE56PXzG98zYm_hY8RXGLI3n-TGk8HV7IxW4Ik-OCnbXYj3T5vwu2q-526wexebp_XK82Ak0BonRkmkKVgNb6PHPe6kxJKZUpyTidk9cKFOaoSFqlnXGqldTkAA0WBgq1YNd_2rmr_khhj-mnnvrquU_9An5zSlo</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Frame Aggregation and Multi-Modal Fusion Framework for Video-Based Person Recognition</title><source>arXiv.org</source><creator>Li, Fangtao ; Wang, Wenzhe ; Liu, Zihe ; Wang, Haoran ; Yan, Chenghao ; Wu, Bin</creator><creatorcontrib>Li, Fangtao ; Wang, Wenzhe ; Liu, Zihe ; Wang, Haoran ; Yan, Chenghao ; Wu, Bin</creatorcontrib><description>Video-based person recognition is challenging due to persons being blocked and blurred, and the variation of shooting angle. Previous research always focused on person recognition on still images, ignoring similarity and continuity between video frames. To tackle the challenges above, we propose a novel Frame Aggregation and Multi-Modal Fusion (FAMF) framework for video-based person recognition, which aggregates face features and incorporates them with multi-modal information to identify persons in videos. For frame aggregation, we propose a novel trainable layer based on NetVLAD (named AttentionVLAD), which takes arbitrary number of features as input and computes a fixed-length aggregation feature based on feature quality. We show that introducing an attention mechanism to NetVLAD can effectively decrease the impact of low-quality frames. For the multi-model information of videos, we propose a Multi-Layer Multi-Modal Attention (MLMA) module to learn the correlation of multi-modality by adaptively updating Gram matrix. Experimental results on iQIYI-VID-2019 dataset show that our framework outperforms other state-of-the-art methods.</description><identifier>DOI: 10.48550/arxiv.2010.09290</identifier><language>eng</language><subject>Computer Science - Computer Vision and Pattern Recognition ; Computer Science - Multimedia</subject><creationdate>2020-10</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2010.09290$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2010.09290$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Li, Fangtao</creatorcontrib><creatorcontrib>Wang, Wenzhe</creatorcontrib><creatorcontrib>Liu, Zihe</creatorcontrib><creatorcontrib>Wang, Haoran</creatorcontrib><creatorcontrib>Yan, Chenghao</creatorcontrib><creatorcontrib>Wu, Bin</creatorcontrib><title>Frame Aggregation and Multi-Modal Fusion Framework for Video-Based Person Recognition</title><description>Video-based person recognition is challenging due to persons being blocked and blurred, and the variation of shooting angle. Previous research always focused on person recognition on still images, ignoring similarity and continuity between video frames. To tackle the challenges above, we propose a novel Frame Aggregation and Multi-Modal Fusion (FAMF) framework for video-based person recognition, which aggregates face features and incorporates them with multi-modal information to identify persons in videos. For frame aggregation, we propose a novel trainable layer based on NetVLAD (named AttentionVLAD), which takes arbitrary number of features as input and computes a fixed-length aggregation feature based on feature quality. We show that introducing an attention mechanism to NetVLAD can effectively decrease the impact of low-quality frames. For the multi-model information of videos, we propose a Multi-Layer Multi-Modal Attention (MLMA) module to learn the correlation of multi-modality by adaptively updating Gram matrix. Experimental results on iQIYI-VID-2019 dataset show that our framework outperforms other state-of-the-art methods.</description><subject>Computer Science - Computer Vision and Pattern Recognition</subject><subject>Computer Science - Multimedia</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj8tOwzAQRb1hgQofwAr_gMs4jh1nWSoCSK1AqLCNJvYkskhj5LQ8_p4msLrS0dWRDmNXEpa51RpuMH2Hz2UGJwBlVsI5e60S7omvui5Rh4cQB46D59tjfwhiGz32vDqOE56PXzG98zYm_hY8RXGLI3n-TGk8HV7IxW4Ik-OCnbXYj3T5vwu2q-526wexebp_XK82Ak0BonRkmkKVgNb6PHPe6kxJKZUpyTidk9cKFOaoSFqlnXGqldTkAA0WBgq1YNd_2rmr_khhj-mnnvrquU_9An5zSlo</recordid><startdate>20201019</startdate><enddate>20201019</enddate><creator>Li, Fangtao</creator><creator>Wang, Wenzhe</creator><creator>Liu, Zihe</creator><creator>Wang, Haoran</creator><creator>Yan, Chenghao</creator><creator>Wu, Bin</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20201019</creationdate><title>Frame Aggregation and Multi-Modal Fusion Framework for Video-Based Person Recognition</title><author>Li, Fangtao ; Wang, Wenzhe ; Liu, Zihe ; Wang, Haoran ; Yan, Chenghao ; Wu, Bin</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a670-9ce6b7390a88d42cd8523111369e6c54ed5303a4a3e1835c6c3f1eb400ba76073</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Computer Science - Computer Vision and Pattern Recognition</topic><topic>Computer Science - Multimedia</topic><toplevel>online_resources</toplevel><creatorcontrib>Li, Fangtao</creatorcontrib><creatorcontrib>Wang, Wenzhe</creatorcontrib><creatorcontrib>Liu, Zihe</creatorcontrib><creatorcontrib>Wang, Haoran</creatorcontrib><creatorcontrib>Yan, Chenghao</creatorcontrib><creatorcontrib>Wu, Bin</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Li, Fangtao</au><au>Wang, Wenzhe</au><au>Liu, Zihe</au><au>Wang, Haoran</au><au>Yan, Chenghao</au><au>Wu, Bin</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Frame Aggregation and Multi-Modal Fusion Framework for Video-Based Person Recognition</atitle><date>2020-10-19</date><risdate>2020</risdate><abstract>Video-based person recognition is challenging due to persons being blocked and blurred, and the variation of shooting angle. Previous research always focused on person recognition on still images, ignoring similarity and continuity between video frames. To tackle the challenges above, we propose a novel Frame Aggregation and Multi-Modal Fusion (FAMF) framework for video-based person recognition, which aggregates face features and incorporates them with multi-modal information to identify persons in videos. For frame aggregation, we propose a novel trainable layer based on NetVLAD (named AttentionVLAD), which takes arbitrary number of features as input and computes a fixed-length aggregation feature based on feature quality. We show that introducing an attention mechanism to NetVLAD can effectively decrease the impact of low-quality frames. For the multi-model information of videos, we propose a Multi-Layer Multi-Modal Attention (MLMA) module to learn the correlation of multi-modality by adaptively updating Gram matrix. Experimental results on iQIYI-VID-2019 dataset show that our framework outperforms other state-of-the-art methods.</abstract><doi>10.48550/arxiv.2010.09290</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2010.09290
ispartof
issn
language eng
recordid cdi_arxiv_primary_2010_09290
source arXiv.org
subjects Computer Science - Computer Vision and Pattern Recognition
Computer Science - Multimedia
title Frame Aggregation and Multi-Modal Fusion Framework for Video-Based Person Recognition
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-28T15%3A58%3A04IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Frame%20Aggregation%20and%20Multi-Modal%20Fusion%20Framework%20for%20Video-Based%20Person%20Recognition&rft.au=Li,%20Fangtao&rft.date=2020-10-19&rft_id=info:doi/10.48550/arxiv.2010.09290&rft_dat=%3Carxiv_GOX%3E2010_09290%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true