计算机与现代化Issue(8)：44-52,9.DOI:10.3969/j.issn.1006-2475.2026.08.006
ECS-YOLO:一种改进的轻量级多菜品检测方法
ECS-YOLO:An Improved Lightweight Method for Multi-dish Detection
摘要
Abstract
In order to better meet the requirement for efficient and accurate multi-dish recognition in smart device environ-ments,this paper proposes ECS-YOLO,a novel architecture based on YOLOv11.The method enhances edge and spatial infor-mation captured by a new backbone network ECC.It developed an improved CFGLU convolutional module to handle complex dish detection scenarios by modifying the original C3K2 structure.The algorithm strengthened occlusion processing capabilities by integrating SEAM into the detection head.Comparative experiments on the kuan-v1 dataset validated ECS-YOLO's effective-ness.Results showed that ECS-YOLO achieved 1.4 percentage points improvements in both Precision and mAP50 compared with YOLOv11,while realizing a lightweight design through 15.4%parameter reduction and 9.4%FLOPs decrease.The study demon-strates that ECS-YOLO maintains superior performance while significantly reducing computational costs.关键词
深度学习/目标检测/菜品检测/轻量化/YOLOKey words
deep learning/object detection/dish recognition/lightweight/YOLO分类
信息技术与安全科学引用本文复制引用
蔡云海,孙晓翠,易法令..ECS-YOLO:一种改进的轻量级多菜品检测方法[J].计算机与现代化,2026,(8):44-52,9.基金项目
广东省重点领域研发计划项目(2022B111102006) （2022B111102006）
广东省中医药管理局科研项目(20232090) （20232090）
