{"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/learning-to-decompose-and-disentangle","title":"Learning to Decompose and Disentangle Representations for Video Prediction","arxiv_id":"1806.04166","date":"2018-06-11","proceeding":"NeurIPS 2018 12","authors":["Jun-Ting Hsieh","Bingbin Liu","De-An Huang","Li Fei-Fei","Juan Carlos Niebles"],"abstract":"Our goal is to predict future video frames given a sequence of input frames.\nDespite large amounts of video data, this remains a challenging task because of\nthe high-dimensionality of video frames. We address this challenge by proposing\nthe Decompositional Disentangled Predictive Auto-Encoder (DDPAE), a framework\nthat combines structured probabilistic models and deep networks to\nautomatically (i) decompose the high-dimensional video that we aim to predict\ninto components, and (ii) disentangle each component to have low-dimensional\ntemporal dynamics that are easier to predict. Crucially, with an appropriately\nspecified generative model of video frames, our DDPAE is able to learn both the\nlatent decomposition and disentanglement without explicit supervision. For the\nMoving MNIST dataset, we show that DDPAE is able to recover the underlying\ncomponents (individual digits) and disentanglement (appearance and location) as\nwe would intuitively do. We further demonstrate that DDPAE can be applied to\nthe Bouncing Balls dataset involving complex interactions between multiple\nobjects to predict the video frame directly from the pixels and recover\nphysical states without explicit supervision.","url_abs":"http://arxiv.org/abs/1806.04166v2","url_pdf":"http://arxiv.org/pdf/1806.04166v2.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":"learning-to-decompose-and-disentangle","repo_url":"https://github.com/jthsieh/DDPAE-video-prediction","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"disentanglement","task_name":"Disentanglement"},{"task_slug":"predict-future-video-frames","task_name":"Predict Future Video Frames"},{"task_slug":"video-prediction","task_name":"Video Prediction"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1806.04166","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}