{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/cdc-convolutional-de-convolutional-networks","title":"CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos","arxiv_id":"1703.01515","date":"2017-03-04","proceeding":"CVPR 2017 7","authors":["Zheng Shou","Jonathan Chan","Alireza Zareian","Kazuyuki Miyazawa","Shih-Fu Chang"],"abstract":"Temporal action localization is an important yet challenging problem. Given a\nlong, untrimmed video consisting of multiple action instances and complex\nbackground contents, we need not only to recognize their action categories, but\nalso to localize the start time and end time of each instance. Many\nstate-of-the-art systems use segment-level classifiers to select and rank\nproposal segments of pre-determined boundaries. However, a desirable model\nshould move beyond segment-level and make dense predictions at a fine\ngranularity in time to determine precise temporal boundaries. To this end, we\ndesign a novel Convolutional-De-Convolutional (CDC) network that places CDC\nfilters on top of 3D ConvNets, which have been shown to be effective for\nabstracting action semantics but reduce the temporal length of the input data.\nThe proposed CDC filter performs the required temporal upsampling and spatial\ndownsampling operations simultaneously to predict actions at the frame-level\ngranularity. It is unique in jointly modeling action semantics in space-time\nand fine-grained temporal dynamics. We train the CDC network in an end-to-end\nmanner efficiently. Our model not only achieves superior performance in\ndetecting actions in every frame, but also significantly boosts the precision\nof localizing temporal boundaries. Finally, the CDC network demonstrates a very\nhigh efficiency with the ability to process 500 frames per second on a single\nGPU server. We will update the camera-ready version and publish the source\ncodes online soon.","url_abs":"http://arxiv.org/abs/1703.01515v2","url_pdf":"http://arxiv.org/pdf/1703.01515v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"cdc-convolutional-de-convolutional-networks","repo_url":"https://bitbucket.org/columbiadvmm/cdc","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"action-localization","task_name":"Action Localization"},{"task_slug":null,"task_name":"GPU"},{"task_slug":"action-recognition","task_name":"Temporal Action Localization"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/temporal-action-localization-on-thumos14","task":"Temporal Action Localization","dataset":"THUMOS’14","model":"CDC","rank_in_archive_order":40,"of":42,"metrics":{"mAP IOU@0.3":"40.1","mAP IOU@0.4":"29.4","mAP IOU@0.5":"23.3","mAP IOU@0.6":"13.1","mAP IOU@0.7":"7.9"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1703.01515","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}