{"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/a-unified-pyramid-recurrent-network-for-video","title":"A Unified Pyramid Recurrent Network for Video Frame Interpolation","arxiv_id":"2211.03456","date":"2022-11-07","proceeding":"CVPR 2023 1","authors":["Xin Jin","Longhai Wu","Jie Chen","Youxin Chen","Jayoon Koo","Cheul-hee Hahm"],"abstract":"Flow-guided synthesis provides a common framework for frame interpolation, where optical flow is estimated to guide the synthesis of intermediate frames between consecutive inputs. In this paper, we present UPR-Net, a novel Unified Pyramid Recurrent Network for frame interpolation. Cast in a flexible pyramid framework, UPR-Net exploits lightweight recurrent modules for both bi-directional flow estimation and intermediate frame synthesis. At each pyramid level, it leverages estimated bi-directional flow to generate forward-warped representations for frame synthesis; across pyramid levels, it enables iterative refinement for both optical flow and intermediate frame. In particular, we show that our iterative synthesis strategy can significantly improve the robustness of frame interpolation on large motion cases. Despite being extremely lightweight (1.7M parameters), our base version of UPR-Net achieves excellent performance on a large range of benchmarks. Code and trained models of our UPR-Net series are available at: https://github.com/srcn-ivl/UPR-Net.","url_abs":"https://arxiv.org/abs/2211.03456v2","url_pdf":"https://arxiv.org/pdf/2211.03456v2.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":"a-unified-pyramid-recurrent-network-for-video","repo_url":"https://github.com/srcn-ivl/upr-net","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"optical-flow-estimation","task_name":"Optical Flow Estimation"},{"task_slug":"video-frame-interpolation","task_name":"Video Frame Interpolation"}],"methods":[{"method_slug":"base","method_name":"BASE"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/video-frame-interpolation-on-msu-video-frame","task":"Video Frame Interpolation","dataset":"MSU Video Frame Interpolation","model":"UPR-Net LARGE","rank_in_archive_order":5,"of":24,"metrics":{"LPIPS":"0.025","MS-SSIM":"0.962","PSNR":"29.73","SSIM":"0.951","VMAF":"71.34"},"uses_additional_data":false},{"leaderboard":"/sota/video-frame-interpolation-on-snu-film-easy","task":"Video Frame Interpolation","dataset":"SNU-FILM (easy)","model":"UPR-Net LARGE","rank_in_archive_order":4,"of":8,"metrics":{"PSNR":"40.44","SSIM":"0.9911"},"uses_additional_data":false},{"leaderboard":"/sota/video-frame-interpolation-on-snu-film-extreme","task":"Video Frame Interpolation","dataset":"SNU-FILM (extreme)","model":"UPR-Net LARGE","rank_in_archive_order":5,"of":8,"metrics":{"PSNR":"25.63","SSIM":"0.8641"},"uses_additional_data":false},{"leaderboard":"/sota/video-frame-interpolation-on-snu-film-hard","task":"Video Frame Interpolation","dataset":"SNU-FILM (hard)","model":"UPR-Net LARGE","rank_in_archive_order":6,"of":8,"metrics":{"PSNR":"30.86","SSIM":"0.9377"},"uses_additional_data":false},{"leaderboard":"/sota/video-frame-interpolation-on-snu-film-medium","task":"Video Frame Interpolation","dataset":"SNU-FILM (medium)","model":"UPR-Net LARGE","rank_in_archive_order":4,"of":8,"metrics":{"PSNR":"36.29","SSIM":"0.9801"},"uses_additional_data":false},{"leaderboard":"/sota/video-frame-interpolation-on-ucf101-1","task":"Video Frame Interpolation","dataset":"UCF101","model":"UPR-Net LARGE","rank_in_archive_order":2,"of":19,"metrics":{"PSNR":"35.47","SSIM":"0.9700"},"uses_additional_data":false},{"leaderboard":"/sota/video-frame-interpolation-on-vimeo90k","task":"Video Frame Interpolation","dataset":"Vimeo90K","model":"UPR-Net LARGE","rank_in_archive_order":4,"of":23,"metrics":{"PSNR":"36.42","SSIM":"0.9815"},"uses_additional_data":false},{"leaderboard":"/sota/video-frame-interpolation-on-x4k1000fps","task":"Video Frame Interpolation","dataset":"X4K1000FPS","model":"UPR-Net large","rank_in_archive_order":8,"of":20,"metrics":{"PSNR":"30.68","SSIM":"0.9086"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2211.03456","atlas_url":"https://app.syntology.ai/?focus=2211.03456","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}