{"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/optical-flow-estimation-using-a-spatial","title":"Optical Flow Estimation using a Spatial Pyramid Network","arxiv_id":"1611.00850","date":"2016-11-03","proceeding":"CVPR 2017 7","authors":["Anurag Ranjan","Michael J. Black"],"abstract":"We learn to compute optical flow by combining a classical spatial-pyramid\nformulation with deep learning. This estimates large motions in a\ncoarse-to-fine approach by warping one image of a pair at each pyramid level by\nthe current flow estimate and computing an update to the flow. Instead of the\nstandard minimization of an objective function at each pyramid level, we train\none deep network per level to compute the flow update. Unlike the recent\nFlowNet approach, the networks do not need to deal with large motions; these\nare dealt with by the pyramid. This has several advantages. First, our Spatial\nPyramid Network (SPyNet) is much simpler and 96% smaller than FlowNet in terms\nof model parameters. This makes it more efficient and appropriate for embedded\napplications. Second, since the flow at each pyramid level is small (< 1\npixel), a convolutional approach applied to pairs of warped images is\nappropriate. Third, unlike FlowNet, the learned convolution filters appear\nsimilar to classical spatio-temporal filters, giving insight into the method\nand how to improve it. Our results are more accurate than FlowNet on most\nstandard benchmarks, suggesting a new direction of combining classical flow\nmethods with deep learning.","url_abs":"http://arxiv.org/abs/1611.00850v2","url_pdf":"http://arxiv.org/pdf/1611.00850v2.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":"optical-flow-estimation-using-a-spatial","repo_url":"https://github.com/anuragranj/spynet","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"torch","reach":{"status":"unanswered"}},{"paper_slug":"optical-flow-estimation-using-a-spatial","repo_url":"https://github.com/GibranBenitez/IPN-hand","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"optical-flow-estimation-using-a-spatial","repo_url":"https://github.com/Yijunmaverick/FlowGrounded-VideoPrediction","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"optical-flow-estimation-using-a-spatial","repo_url":"https://github.com/guanfuchen/video_obj","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"optical-flow-estimation-using-a-spatial","repo_url":"https://github.com/philferriere/tfoptflow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"optical-flow-estimation-using-a-spatial","repo_url":"https://github.com/rickyHong/tfoptflow-repl","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"optical-flow-estimation-using-a-spatial","repo_url":"https://github.com/sniklaus/pytorch-spynet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"optical-flow-estimation-using-a-spatial","repo_url":"https://github.com/Guillem96/spynet-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"dense-pixel-correspondence-estimation","task_name":"Dense Pixel Correspondence Estimation"},{"task_slug":"optical-flow-estimation","task_name":"Optical Flow Estimation"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/dense-pixel-correspondence-estimation-on","task":"Dense Pixel Correspondence Estimation","dataset":"HPatches","model":"SPyNet","rank_in_archive_order":8,"of":8,"metrics":{"Viewpoint I AEPE":"36.94","Viewpoint II AEPE":"50.92","Viewpoint III AEPE":"54.29","Viewpoint IV AEPE":"62.60","Viewpoint V AEPE":"72.57"},"uses_additional_data":false},{"leaderboard":"/sota/optical-flow-estimation-on-sintel-clean","task":"Optical Flow Estimation","dataset":"Sintel-clean","model":"Spynet","rank_in_archive_order":29,"of":29,"metrics":{"Average End-Point Error":"6.64"},"uses_additional_data":false},{"leaderboard":"/sota/optical-flow-estimation-on-sintel-final","task":"Optical Flow Estimation","dataset":"Sintel-final","model":"Spynet","rank_in_archive_order":28,"of":28,"metrics":{"Average End-Point Error":"8.36"},"uses_additional_data":false},{"leaderboard":"/sota/optical-flow-estimation-on-spring","task":"Optical Flow Estimation","dataset":"Spring","model":"SPyNet","rank_in_archive_order":10,"of":11,"metrics":{"1px total":"29.963"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1611.00850","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}