{"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/curvelane-nas-unifying-lane-sensitive","title":"CurveLane-NAS: Unifying Lane-Sensitive Architecture Search and Adaptive Point Blending","arxiv_id":"2007.12147","date":"2020-07-23","proceeding":"ECCV 2020 8","authors":["Hang Xu","Shaoju Wang","Xinyue Cai","Wei zhang","Xiaodan Liang","Zhenguo Li"],"abstract":"We address the curve lane detection problem which poses more realistic challenges than conventional lane detection for better facilitating modern assisted/autonomous driving systems. Current hand-designed lane detection methods are not robust enough to capture the curve lanes especially the remote parts due to the lack of modeling both long-range contextual information and detailed curve trajectory. In this paper, we propose a novel lane-sensitive architecture search framework named CurveLane-NAS to automatically capture both long-ranged coherent and accurate short-range curve information while unifying both architecture search and post-processing on curve lane predictions via point blending. It consists of three search modules: a) a feature fusion search module to find a better fusion of the local and global context for multi-level hierarchy features; b) an elastic backbone search module to explore an efficient feature extractor with good semantics and latency; c) an adaptive point blending module to search a multi-level post-processing refinement strategy to combine multi-scale head prediction. The unified framework ensures lane-sensitive predictions by the mutual guidance between NAS and adaptive point blending. Furthermore, we also steer forward to release a more challenging benchmark named CurveLanes for addressing the most difficult curve lanes. It consists of 150K images with 680K labels.The new dataset can be downloaded at github.com/xbjxh/CurveLanes (already anonymized for this submission). Experiments on the new CurveLanes show that the SOTA lane detection methods suffer substantial performance drop while our model can still reach an 80+% F1-score. Extensive experiments on traditional lane benchmarks such as CULane also demonstrate the superiority of our CurveLane-NAS, e.g. achieving a new SOTA 74.8% F1-score on CULane.","url_abs":"https://arxiv.org/abs/2007.12147v1","url_pdf":"https://arxiv.org/pdf/2007.12147v1.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":"curvelane-nas-unifying-lane-sensitive","repo_url":"https://github.com/huawei-noah/vega","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"autonomous-driving","task_name":"Autonomous Driving"},{"task_slug":"lane-detection","task_name":"Lane Detection"}],"methods":[],"datasets_introduced":[{"slug":"curvelanes","name":"CurveLanes","full_name":""}],"methods_introduced":[],"results":[{"leaderboard":"/sota/lane-detection-on-culane","task":"Lane Detection","dataset":"CULane","model":"CurveLane-L","rank_in_archive_order":47,"of":63,"metrics":{"F1 score":"74.8"},"uses_additional_data":false},{"leaderboard":"/sota/lane-detection-on-culane","task":"Lane Detection","dataset":"CULane","model":"CurveLane-M","rank_in_archive_order":54,"of":63,"metrics":{"F1 score":"73.5"},"uses_additional_data":false},{"leaderboard":"/sota/lane-detection-on-culane","task":"Lane Detection","dataset":"CULane","model":"CurveLane-S","rank_in_archive_order":59,"of":63,"metrics":{"F1 score":"71.4"},"uses_additional_data":false},{"leaderboard":"/sota/lane-detection-on-curvelanes","task":"Lane Detection","dataset":"CurveLanes","model":"CurveLane-L","rank_in_archive_order":13,"of":19,"metrics":{"F1 score":"82.29","GFLOPs":"20.7","Precision":"91.11","Recall":"75.03"},"uses_additional_data":false},{"leaderboard":"/sota/lane-detection-on-curvelanes","task":"Lane Detection","dataset":"CurveLanes","model":"CurveLane-M","rank_in_archive_order":14,"of":19,"metrics":{"F1 score":"81.8","GFLOPs":"11.6","Precision":"93.49","Recall":"72.71"},"uses_additional_data":false},{"leaderboard":"/sota/lane-detection-on-curvelanes","task":"Lane Detection","dataset":"CurveLanes","model":"CurveLane-S","rank_in_archive_order":15,"of":19,"metrics":{"F1 score":"81.12","GFLOPs":"7.4","Precision":"93.58","Recall":"71.59"},"uses_additional_data":false},{"leaderboard":"/sota/lane-detection-on-curvelanes","task":"Lane Detection","dataset":"CurveLanes","model":"PointLaneNet","rank_in_archive_order":16,"of":19,"metrics":{"F1 score":"78.47","GFLOPs":"14.8","Precision":"86.33","Recall":"72.91"},"uses_additional_data":false},{"leaderboard":"/sota/lane-detection-on-curvelanes","task":"Lane Detection","dataset":"CurveLanes","model":"SCNN","rank_in_archive_order":17,"of":19,"metrics":{"F1 score":"65.02","GFLOPs":"328.4","Precision":"76.13","Recall":"56.74"},"uses_additional_data":false},{"leaderboard":"/sota/lane-detection-on-curvelanes","task":"Lane Detection","dataset":"CurveLanes","model":"Enet-SAD","rank_in_archive_order":18,"of":19,"metrics":{"F1 score":"50.31","GFLOPs":"3.9","Precision":"63.6","Recall":"41.6"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2007.12147","atlas_url":"https://app.syntology.ai/?focus=2007.12147","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}