{"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/bevdepth-acquisition-of-reliable-depth-for","title":"BEVDepth: Acquisition of Reliable Depth for Multi-view 3D Object Detection","arxiv_id":"2206.10092","date":"2022-06-21","proceeding":null,"authors":["Yinhao Li","Zheng Ge","Guanyi Yu","Jinrong Yang","Zengran Wang","Yukang Shi","Jianjian Sun","Zeming Li"],"abstract":"In this research, we propose a new 3D object detector with a trustworthy depth estimation, dubbed BEVDepth, for camera-based Bird's-Eye-View (BEV) 3D object detection. Our work is based on a key observation -- depth estimation in recent approaches is surprisingly inadequate given the fact that depth is essential to camera 3D detection. Our BEVDepth resolves this by leveraging explicit depth supervision. A camera-awareness depth estimation module is also introduced to facilitate the depth predicting capability. Besides, we design a novel Depth Refinement Module to counter the side effects carried by imprecise feature unprojection. Aided by customized Efficient Voxel Pooling and multi-frame mechanism, BEVDepth achieves the new state-of-the-art 60.9% NDS on the challenging nuScenes test set while maintaining high efficiency. For the first time, the NDS score of a camera model reaches 60%.","url_abs":"https://arxiv.org/abs/2206.10092v2","url_pdf":"https://arxiv.org/pdf/2206.10092v2.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":"bevdepth-acquisition-of-reliable-depth-for","repo_url":"https://github.com/megvii-basedetection/bevdepth","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"bevdepth-acquisition-of-reliable-depth-for","repo_url":"https://github.com/ZRandomize/MatrixVT","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"3d-object-detection","task_name":"3D Object Detection"},{"task_slug":"depth-estimation","task_name":"Depth Estimation"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"robust-camera-only-3d-object-detection","task_name":"Robust Camera Only 3D Object Detection"}],"methods":[{"method_slug":"speed","method_name":"SPEED"},{"method_slug":"test","method_name":"Test"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/3d-object-detection-on-dair-v2x-i","task":"3D Object Detection","dataset":"DAIR-V2X-I","model":"BEVDepth","rank_in_archive_order":4,"of":9,"metrics":{"AP|R40(easy)":"75.7","AP|R40(hard)":"63.7","AP|R40(moderate)":"63.6"},"uses_additional_data":false},{"leaderboard":"/sota/3d-object-detection-on-rope3d","task":"3D Object Detection","dataset":"Rope3D","model":"BEVDepth","rank_in_archive_order":4,"of":8,"metrics":{"AP@0.7":"42.56"},"uses_additional_data":false},{"leaderboard":"/sota/3d-object-detection-on-nuscenes-camera-only","task":"3D Object Detection","dataset":"nuScenes Camera Only","model":"BEVDepth-pure","rank_in_archive_order":13,"of":19,"metrics":{"Future Frame":"false","NDS":"60.9"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2206.10092","atlas_url":"https://app.syntology.ai/?focus=2206.10092","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}