{"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/comedian-self-supervised-learning-and","title":"COMEDIAN: Self-Supervised Learning and Knowledge Distillation for Action Spotting using Transformers","arxiv_id":"2309.01270","date":"2023-09-03","proceeding":null,"authors":["Julien Denize","Mykola Liashuha","Jaonary Rabarisoa","Astrid Orcesi","Romain Hérault"],"abstract":"We present COMEDIAN, a novel pipeline to initialize spatiotemporal transformers for action spotting, which involves self-supervised learning and knowledge distillation. Action spotting is a timestamp-level temporal action detection task. Our pipeline consists of three steps, with two initialization stages. First, we perform self-supervised initialization of a spatial transformer using short videos as input. Additionally, we initialize a temporal transformer that enhances the spatial transformer's outputs with global context through knowledge distillation from a pre-computed feature bank aligned with each short video segment. In the final step, we fine-tune the transformers to the action spotting task. The experiments, conducted on the SoccerNet-v2 dataset, demonstrate state-of-the-art performance and validate the effectiveness of COMEDIAN's pretraining paradigm. Our results highlight several advantages of our pretraining pipeline, including improved performance and faster convergence compared to non-pretrained models.","url_abs":"https://arxiv.org/abs/2309.01270v2","url_pdf":"https://arxiv.org/pdf/2309.01270v2.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":"comedian-self-supervised-learning-and","repo_url":"https://github.com/juliendenize/eztorch","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"action-detection","task_name":"Action Detection"},{"task_slug":"action-spotting","task_name":"Action Spotting"},{"task_slug":"knowledge-distillation","task_name":"Knowledge Distillation"},{"task_slug":"self-supervised-learning","task_name":"Self-Supervised Learning"}],"methods":[{"method_slug":"knowledge-distillation","method_name":"Knowledge Distillation"},{"method_slug":"spatial-transformer","method_name":"Spatial Transformer"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/action-spotting-on-soccernet-v2","task":"Action Spotting","dataset":"SoccerNet-v2","model":"COMEDIAN (ViSwin T ens.)","rank_in_archive_order":1,"of":10,"metrics":{"Average-mAP":"77.6","Tight Average-mAP":"73.1"},"uses_additional_data":false},{"leaderboard":"/sota/action-spotting-on-soccernet-v2","task":"Action Spotting","dataset":"SoccerNet-v2","model":"COMEDIAN (ViViT T ens.)","rank_in_archive_order":2,"of":10,"metrics":{"Average-mAP":"77.1","Tight Average-mAP":"72.0"},"uses_additional_data":false},{"leaderboard":"/sota/action-spotting-on-soccernet-v2","task":"Action Spotting","dataset":"SoccerNet-v2","model":"COMEDIAN (ViSwin T)","rank_in_archive_order":3,"of":10,"metrics":{"Average-mAP":"76.6","Tight Average-mAP":"71.6"},"uses_additional_data":false},{"leaderboard":"/sota/action-spotting-on-soccernet-v2","task":"Action Spotting","dataset":"SoccerNet-v2","model":"COMEDIAN (ViViT T)","rank_in_archive_order":4,"of":10,"metrics":{"Average-mAP":"76.1","Tight Average-mAP":"70.7"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2309.01270","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}