Recognizing actions in videos is a challenging task as video is an information-intensive media with complex variations. Most existing methods have treated video as a flat data sequence while ignoring the intrinsic hierarchical structure of the video content. In particular, an action may span different granularities in this hierarchy including, from small to large, a single \\emph{frame}, consecutive frames (\\emph{motion}), a short \\emph{clip}, and the entire \\emph{video}. In this paper, we present a novel framework to boost action recognition by learning a deep spatio-temporal video representation at hierarchical multi-granularity. Specifically, we model each granularity as a single stream by 2D (for \\emph{frame} and \\emph{motion} streams) or 3D (for \\emph{clip} and \\emph{video} streams) convolutional neural networks (CNNs). The framework therefore consists of multi-stream 2D or 3D CNNs to learn both the spatial and temporal representations. Furthermore, we employ the Long Short-Term Memory (LSTM) networks on the \\emph{frame}, \\emph{motion}, and \\emph{clip} streams to exploit long-term temporal dynamics. With a \\emph{softmax} layer on the top of each stream, the classification scores can be predicted from all the streams, followed by a novel fusion scheme based on the multi-granular score distribution. Our networks are learned in an end-to-end fashion. On two video action benchmarks of UCF101 and HMDB51, our framework achieves promising performance compared with the state-of-the-art.<\/p>\n","protected":false},"excerpt":{"rendered":"
Recognizing actions in videos is a challenging task as video is an information-intensive media with complex variations. Most existing methods have treated video as a flat data sequence while ignoring the intrinsic hierarchical structure of the video content. In particular, an action may span different granularities in this hierarchy including, from small to large, a […]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-author-ordering":null,"msr_publishername":"","msr_publisher_other":"","msr_booktitle":"","msr_chapter":"","msr_edition":"ACM International Conference in Multimedia Retrieval (ICMR)","msr_editors":"","msr_how_published":"","msr_isbn":"","msr_issue":"","msr_journal":"","msr_number":"","msr_organization":"","msr_pages_string":"","msr_page_range_start":"","msr_page_range_end":"","msr_series":"","msr_volume":"","msr_copyright":"","msr_conference_name":"ACM International Conference in Multimedia Retrieval (ICMR)","msr_doi":"","msr_arxiv_id":"","msr_s2_paper_id":"","msr_mag_id":"","msr_pubmed_id":"","msr_other_authors":"","msr_other_contributors":"","msr_speaker":"","msr_award":"Best Paper Finalist (the first two authors made equal contributions)","msr_affiliation":"","msr_institution":"","msr_host":"","msr_version":"","msr_duration":"","msr_original_fields_of_study":"","msr_release_tracker_id":"","msr_s2_match_type":"","msr_citation_count_updated":"","msr_published_date":"2016-06-04","msr_highlight_text":"","msr_notes":"Best Paper Finalist (the first two authors made equal contributions)","msr_longbiography":"","msr_publicationurl":"","msr_external_url":"","msr_secondary_video_url":"","msr_conference_url":"","msr_journal_url":"","msr_s2_pdf_url":"","msr_year":0,"msr_citation_count":0,"msr_influential_citations":0,"msr_reference_count":0,"msr_s2_match_confidence":0,"msr_microsoftintellectualproperty":true,"msr_s2_open_access":false,"msr_s2_author_ids":[],"msr_pub_ids":[],"msr_hide_image_in_river":0,"footnotes":""},"msr-research-highlight":[],"research-area":[13562,13551],"msr-publication-type":[193716],"msr-publisher":[],"msr-focus-area":[],"msr-locale":[268875],"msr-post-option":[],"msr-field-of-study":[],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-271185","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-computer-vision","msr-research-area-graphics-and-multimedia","msr-locale-en_us"],"msr_publishername":"","msr_edition":"ACM International Conference in Multimedia Retrieval (ICMR)","msr_affiliation":"","msr_published_date":"2016-06-04","msr_host":"","msr_duration":"","msr_version":"","msr_speaker":"","msr_other_contributors":"","msr_booktitle":"","msr_pages_string":"","msr_chapter":"","msr_isbn":"","msr_journal":"","msr_volume":"","msr_number":"","msr_editors":"","msr_series":"","msr_issue":"","msr_organization":"","msr_how_published":"","msr_notes":"Best Paper Finalist (the first two authors made equal contributions)","msr_highlight_text":"","msr_release_tracker_id":"","msr_original_fields_of_study":"","msr_download_urls":"","msr_external_url":"","msr_secondary_video_url":"","msr_longbiography":"","msr_microsoftintellectualproperty":1,"msr_main_download":"271191","msr_publicationurl":"","msr_doi":"","msr_publication_uploader":[{"type":"file","title":"icmr038-liA","viewUrl":"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2016\/08\/icmr038-liA.pdf","id":271191,"label_id":0}],"msr_related_uploader":"","msr_citation_count":0,"msr_citation_count_updated":"","msr_s2_paper_id":"","msr_influential_citations":0,"msr_reference_count":0,"msr_arxiv_id":"","msr_s2_author_ids":[],"msr_s2_open_access":false,"msr_s2_pdf_url":null,"msr_attachments":[],"msr-author-ordering":[{"type":"text","value":"Qing Li","user_id":0,"rest_url":false},{"type":"text","value":"Zhaofan Qiu","user_id":0,"rest_url":false},{"type":"user_nicename","value":"tiyao","user_id":34069,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=tiyao"},{"type":"user_nicename","value":"tmei","user_id":34188,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=tmei"},{"type":"user_nicename","value":"yongrui","user_id":35040,"rest_url":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=yongrui"},{"type":"text","value":"Jiebo Luo","user_id":0,"rest_url":false}],"msr_impact_theme":[],"msr_research_lab":[199560],"msr_event":[],"msr_group":[144916],"msr_project":[239357],"publication":[],"video":[],"msr-tool":[],"msr_publication_type":"inproceedings","related_content":{"projects":[{"ID":239357,"post_title":"Video Analysis","post_name":"video-analytics","post_type":"msr-project","post_date":"2016-06-16 19:35:23","post_modified":"2017-10-07 21:38:55","post_status":"publish","permalink":"https:\/\/www.microsoft.com\/en-us\/research\/project\/video-analytics\/","post_excerpt":"Video has become ubiquitous on the Internet, broadcasting channels, as well as that captured by personal devices. This has encouraged the development of advanced techniques to analyze the semantic video content for a wide variety of applications, such as video representation learning [CVPR 2017], video highlight detection [CVPR 2016], video summarization, object detection, action recognition [CVPR 2016, ICMR 2016], semantic segmentation, and so on. Highlight detection The emergence of wearable devices such as portable cameras…","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/239357"}]}}]},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/271185","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":2,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/271185\/revisions"}],"predecessor-version":[{"id":532521,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/271185\/revisions\/532521"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=271185"}],"wp:term":[{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=271185"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=271185"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=271185"},{"taxonomy":"msr-publisher","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-publisher?post=271185"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=271185"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=271185"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=271185"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=271185"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=271185"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=271185"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=271185"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=271185"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}