{"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/towards-accurate-generative-models-of-video-a","title":"Towards Accurate Generative Models of Video: A New Metric & Challenges","arxiv_id":"1812.01717","date":"2018-12-03","proceeding":null,"authors":["Thomas Unterthiner","Sjoerd van Steenkiste","Karol Kurach","Raphael Marinier","Marcin Michalski","Sylvain Gelly"],"abstract":"Recent advances in deep generative models have lead to remarkable progress in\nsynthesizing high quality images. Following their successful application in\nimage processing and representation learning, an important next step is to\nconsider videos. Learning generative models of video is a much harder task,\nrequiring a model to capture the temporal dynamics of a scene, in addition to\nthe visual presentation of objects. While recent attempts at formulating\ngenerative models of video have had some success, current progress is hampered\nby (1) the lack of qualitative metrics that consider visual quality, temporal\ncoherence, and diversity of samples, and (2) the wide gap between purely\nsynthetic video data sets and challenging real-world data sets in terms of\ncomplexity. To this extent we propose Fr\\'{e}chet Video Distance (FVD), a new\nmetric for generative models of video, and StarCraft 2 Videos (SCV), a\nbenchmark of game play from custom starcraft 2 scenarios that challenge the\ncurrent capabilities of generative models of video. We contribute a large-scale\nhuman study, which confirms that FVD correlates well with qualitative human\njudgment of generated videos, and provide initial benchmark results on SCV.","url_abs":"http://arxiv.org/abs/1812.01717v2","url_pdf":"http://arxiv.org/pdf/1812.01717v2.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":"towards-accurate-generative-models-of-video-a","repo_url":"https://github.com/nvlabs/long-video-gan","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"towards-accurate-generative-models-of-video-a","repo_url":"https://github.com/universome/stylegan-v","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"towards-accurate-generative-models-of-video-a","repo_url":"https://github.com/wilson1yan/VideoGPT","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"diversity","task_name":"Diversity"},{"task_slug":"representation-learning","task_name":"Representation Learning"},{"task_slug":"starcraft","task_name":"Starcraft"},{"task_slug":"video-prediction","task_name":"Video Prediction"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1812.01717","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}