[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-2853":3},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":8,"source_name":9,"source_url":8,"published_at":10,"category":11,"cover_url":8,"hotness":12,"is_selected":13,"score":14,"score_detail":15,"sources":23,"tags":25,"view_count":31,"doi":8,"paper":32,"created_at":41},2853,"面向资源受限的茶芽采摘机器人的轻量化视觉感知:COS-DETR压缩感知系统——MDPI Processes","https:\u002F\u002Fwww.mdpi.com\u002F2227-9717\u002F14\u002F18\u002F2963","论文开发了COS-DETR,基于RT-DETR融合Faster CGLU块、OmniKernel模块(含FSAM)和SPDConv。还评估了结合层自适应幅度剪枝(LAMP)、通道剪枝与Mimic+Linear知识蒸馏的压缩流水线。未压缩模型在评估检测器中AP达到87.1%最高值,经剪枝和蒸馏后,在NVIDIA Jetson Orin NX上达到85.7%和58.1 FPS,计算成本降低37.3%,AP仅降低1.4个百分点。15次静态田间试验中,该检测器与立体相机和Delta机械臂集成,绝对定位误差低于5 mm。来自西华大学。",null,"MDPI Processes","2026-09-13T00:00:00Z","论文",10,false,75,{"impact":16,"substance":17,"depth":18,"authority":19,"freshness":20,"relevant":21,"comment":22},16,22,18,13,6,1,"面向茶芽采摘机器人的轻量化视觉感知研究，方法新颖、田间验证扎实，对农业机器人落地有参考价值。",[24],{"name":9,"url":6},[26,27,28,29,30],"智慧农业","农业人工智能","机器人","茶叶采摘","模型轻量化",0,{"doi":8,"openalex_id":8,"authors":33,"venue":8,"cited_by_count":31,"oa_url":8,"card":34,"direction":38,"ingested_from":40},[],{"tldr":35,"method":36,"finding":37,"direction":38,"opportunity":39},"提出COS-DETR轻量化视觉感知系统，用于资源受限的茶芽采摘机器人。","基于RT-DETR融合Faster CGLU、OmniKernel与SPDCon","压缩后AP达85.7%、58.1 FPS，算力降37.3%，定位误差低于5 mm。","农业人工智能与决策模型","可探索多作物通用轻量检测框架及动态剪枝在田间实时采摘中的泛化性。","agent","2026-09-18T00:03:30.679420Z"]