{"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-video-object-segmentation-with","title":"Learning Video Object Segmentation with Visual Memory","arxiv_id":"1704.05737","date":"2017-04-19","proceeding":"ICCV 2017 10","authors":["Pavel Tokmakov","Karteek Alahari","Cordelia Schmid"],"abstract":"This paper addresses the task of segmenting moving objects in unconstrained\nvideos. We introduce a novel two-stream neural network with an explicit memory\nmodule to achieve this. The two streams of the network encode spatial and\ntemporal features in a video sequence respectively, while the memory module\ncaptures the evolution of objects over time. The module to build a \"visual\nmemory\" in video, i.e., a joint representation of all the video frames, is\nrealized with a convolutional recurrent unit learned from a small number of\ntraining video sequences. Given a video frame as input, our approach assigns\neach pixel an object or background label based on the learned spatio-temporal\nfeatures as well as the \"visual memory\" specific to the video, acquired\nautomatically without any manually-annotated frames. The visual memory is\nimplemented with convolutional gated recurrent units, which allows to propagate\nspatial information over time. We evaluate our method extensively on two\nbenchmarks, DAVIS and Freiburg-Berkeley motion segmentation datasets, and show\nstate-of-the-art results. For example, our approach outperforms the top method\non the DAVIS dataset by nearly 6%. We also provide an extensive ablative\nanalysis to investigate the influence of each component in the proposed\nframework.","url_abs":"http://arxiv.org/abs/1704.05737v2","url_pdf":"http://arxiv.org/pdf/1704.05737v2.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":[],"tasks":[{"task_slug":"motion-segmentation","task_name":"Motion Segmentation"},{"task_slug":"object","task_name":"Object"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"unsupervised-video-object-segmentation","task_name":"Unsupervised Video Object Segmentation"},{"task_slug":"video-object-segmentation","task_name":"Video Object Segmentation"},{"task_slug":"video-semantic-segmentation","task_name":"Video Semantic Segmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/unsupervised-video-object-segmentation-on-3","task":"Unsupervised Video Object Segmentation","dataset":"SegTrack v2","model":"LVO","rank_in_archive_order":3,"of":4,"metrics":{"Mean IoU":"57.3"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1704.05737","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}