{"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/asynchronous-temporal-fields-for-action","title":"Asynchronous Temporal Fields for Action Recognition","arxiv_id":"1612.06371","date":"2016-12-19","proceeding":"CVPR 2017 7","authors":["Gunnar A. Sigurdsson","Santosh Divvala","Ali Farhadi","Abhinav Gupta"],"abstract":"Actions are more than just movements and trajectories: we cook to eat and we\nhold a cup to drink from it. A thorough understanding of videos requires going\nbeyond appearance modeling and necessitates reasoning about the sequence of\nactivities, as well as the higher-level constructs such as intentions. But how\ndo we model and reason about these? We propose a fully-connected temporal CRF\nmodel for reasoning over various aspects of activities that includes objects,\nactions, and intentions, where the potentials are predicted by a deep network.\nEnd-to-end training of such structured models is a challenging endeavor: For\ninference and learning we need to construct mini-batches consisting of whole\nvideos, leading to mini-batches with only a few videos. This causes\nhigh-correlation between data points leading to breakdown of the backprop\nalgorithm. To address this challenge, we present an asynchronous variational\ninference method that allows efficient end-to-end training. Our method achieves\na classification mAP of 22.4% on the Charades benchmark, outperforming the\nstate-of-the-art (17.2% mAP), and offers equal gains on the task of temporal\nlocalization.","url_abs":"http://arxiv.org/abs/1612.06371v2","url_pdf":"http://arxiv.org/pdf/1612.06371v2.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":"asynchronous-temporal-fields-for-action","repo_url":"https://github.com/gsig/temporal-fields","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"torch","reach":{"status":"ok"}},{"paper_slug":"asynchronous-temporal-fields-for-action","repo_url":"https://github.com/gsig/charades-algorithms","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"torch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"action-classification","task_name":"Action Classification"},{"task_slug":"action-recognition-in-videos","task_name":"Action Recognition"},{"task_slug":"action-recognition","task_name":"Temporal Action Localization"},{"task_slug":"temporal-localization","task_name":"Temporal Localization"},{"task_slug":"variational-inference","task_name":"Variational Inference"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/action-classification-on-charades","task":"Action Classification","dataset":"Charades","model":"Asyn-TF","rank_in_archive_order":45,"of":49,"metrics":{"MAP":"22.4"},"uses_additional_data":true},{"leaderboard":"/sota/action-detection-on-charades","task":"Action Detection","dataset":"Charades","model":"Sigurdsson et al.","rank_in_archive_order":16,"of":16,"metrics":{"mAP":"9.6"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1612.06371","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}