Atlanta scaled layouts from non-central panoramas
作者：
Highlights：
• We propose two new geometrical solvers for layout recovery from non-central panorama.
• First work that use deep learning with non-central projection systems.
• We extract the layout scale from a single panorama without extra measurements.
• Occlusion management in Manhattan and Atlanta environments from a single image.
• We also present the first data-set of non-central panoramas in the state of the art.
摘要
•We propose two new geometrical solvers for layout recovery from non-central panorama.•First work that use deep learning with non-central projection systems.•We extract the layout scale from a single panorama without extra measurements.•Occlusion management in Manhattan and Atlanta environments from a single image.•We also present the first data-set of non-central panoramas in the state of the art.
论文关键词：Omnidirectional vision,3D vision,Non-central cameras,Layout recovery,Scene understanding
论文评审过程：Received 23 June 2021, Revised 30 March 2022, Accepted 23 April 2022, Available online 25 April 2022, Version of Record 1 May 2022.
论文官网地址：https://doi.org/10.1016/j.patcog.2022.108740
