{"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-blind-video-temporal-consistency","title":"Learning Blind Video Temporal Consistency","arxiv_id":"1808.00449","date":"2018-08-01","proceeding":"ECCV 2018 9","authors":["Wei-Sheng Lai","Jia-Bin Huang","Oliver Wang","Eli Shechtman","Ersin Yumer","Ming-Hsuan Yang"],"abstract":"Applying image processing algorithms independently to each frame of a video\noften leads to undesired inconsistent results over time. Developing temporally\nconsistent video-based extensions, however, requires domain knowledge for\nindividual tasks and is unable to generalize to other applications. In this\npaper, we present an efficient end-to-end approach based on deep recurrent\nnetwork for enforcing temporal consistency in a video. Our method takes the\noriginal unprocessed and per-frame processed videos as inputs to produce a\ntemporally consistent video. Consequently, our approach is agnostic to specific\nimage processing algorithms applied on the original video. We train the\nproposed network by minimizing both short-term and long-term temporal losses as\nwell as the perceptual loss to strike a balance between temporal stability and\nperceptual similarity with the processed frames. At test time, our model does\nnot require computing optical flow and thus achieves real-time speed even for\nhigh-resolution videos. We show that our single model can handle multiple and\nunseen tasks, including but not limited to artistic style transfer,\nenhancement, colorization, image-to-image translation and intrinsic image\ndecomposition. Extensive objective evaluation and subject study demonstrate\nthat the proposed approach performs favorably against the state-of-the-art\nmethods on various types of videos.","url_abs":"http://arxiv.org/abs/1808.00449v1","url_pdf":"http://arxiv.org/pdf/1808.00449v1.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-blind-video-temporal-consistency","repo_url":"https://github.com/phoenix104104/fast_blind_video_consistency","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"colorization","task_name":"Colorization"},{"task_slug":"image-to-image-translation","task_name":"Image-to-Image Translation"},{"task_slug":"intrinsic-image-decomposition","task_name":"Intrinsic Image Decomposition"},{"task_slug":"optical-flow-estimation","task_name":"Optical Flow Estimation"},{"task_slug":"style-transfer","task_name":"Style Transfer"},{"task_slug":"video-temporal-consistency","task_name":"Video Temporal Consistency"}],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1808.00449","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}