{"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/geometric-loss-functions-for-camera-pose","title":"Geometric Loss Functions for Camera Pose Regression with Deep Learning","arxiv_id":"1704.00390","date":"2017-04-02","proceeding":"CVPR 2017 7","authors":["Alex Kendall","Roberto Cipolla"],"abstract":"Deep learning has shown to be effective for robust and real-time monocular\nimage relocalisation. In particular, PoseNet is a deep convolutional neural\nnetwork which learns to regress the 6-DOF camera pose from a single image. It\nlearns to localize using high level features and is robust to difficult\nlighting, motion blur and unknown camera intrinsics, where point based SIFT\nregistration fails. However, it was trained using a naive loss function, with\nhyper-parameters which require expensive tuning. In this paper, we give the\nproblem a more fundamental theoretical treatment. We explore a number of novel\nloss functions for learning camera pose which are based on geometry and scene\nreprojection error. Additionally we show how to automatically learn an optimal\nweighting to simultaneously regress position and orientation. By leveraging\ngeometry, we demonstrate that our technique significantly improves PoseNet's\nperformance across datasets ranging from indoor rooms to a small city.","url_abs":"http://arxiv.org/abs/1704.00390v2","url_pdf":"http://arxiv.org/pdf/1704.00390v2.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":"geometric-loss-functions-for-camera-pose","repo_url":"https://github.com/imanlab/deep_movement_primitives","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"deep-learning","task_name":"Deep Learning"},{"task_slug":null,"task_name":"Position"},{"task_slug":"regression-1","task_name":"regression"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1704.00390","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}