{"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/unflow-unsupervised-learning-of-optical-flow","title":"UnFlow: Unsupervised Learning of Optical Flow with a Bidirectional Census Loss","arxiv_id":"1711.07837","date":"2017-11-21","proceeding":null,"authors":["Simon Meister","Junhwa Hur","Stefan Roth"],"abstract":"In the era of end-to-end deep learning, many advances in computer vision are\ndriven by large amounts of labeled data. In the optical flow setting, however,\nobtaining dense per-pixel ground truth for real scenes is difficult and thus\nsuch data is rare. Therefore, recent end-to-end convolutional networks for\noptical flow rely on synthetic datasets for supervision, but the domain\nmismatch between training and test scenarios continues to be a challenge.\nInspired by classical energy-based optical flow methods, we design an\nunsupervised loss based on occlusion-aware bidirectional flow estimation and\nthe robust census transform to circumvent the need for ground truth flow. On\nthe KITTI benchmarks, our unsupervised approach outperforms previous\nunsupervised deep networks by a large margin, and is even more accurate than\nsimilar supervised methods trained on synthetic datasets alone. By optionally\nfine-tuning on the KITTI training data, our method achieves competitive optical\nflow accuracy on the KITTI 2012 and 2015 benchmarks, thus in addition enabling\ngeneric pre-training of supervised networks for datasets with limited amounts\nof ground truth.","url_abs":"http://arxiv.org/abs/1711.07837v1","url_pdf":"http://arxiv.org/pdf/1711.07837v1.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":"unflow-unsupervised-learning-of-optical-flow","repo_url":"https://github.com/simonmeister/UnFlow","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"unflow-unsupervised-learning-of-optical-flow","repo_url":"https://github.com/sniklaus/pytorch-unflow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"optical-flow-estimation","task_name":"Optical Flow Estimation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1711.07837","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}