Authors

Abstract

Egg external defects reduce commercial value and may compromise product safety. This study proposes YOLOv11n-DAL, a lightweight detector integrating C3k2_DCNv4, ADown, and Detect_LSDECD for conveyor-belt egg inspection. A private dataset of 1,770 images covering three classes labelled damage, dirty, and good was constructed and expanded to 5,310 images. YOLOv11n-DAL achieved 95.2% precision, 94.5% recall, 94.8% F1-score, 96.7% mAP@0.5, and 95.8% mAP@0.5:0.95. The model contained 1.7 M parameters, required 4.8 GFLOPs, and occupied 3.8 MB, demonstrating its potential for lightweight egg-sorting deployment.

Abstract in Chinese

鸡蛋外观缺陷会降低商品价值并可能影响产品安全。本文提出轻量化检测模型 YOLOv11n-DAL，将 C3k2_DCNv4、ADown 和 Detect_LSDECD 融合用于传送带场景下的鸡蛋检测。构建包含 damage、dirty 和 good 三类的 1770 张私有图像数据集，并增强至 5310 张。模型的 Precision、Recall、F1-score、mAP@0.5 和 mAP@0.5:0.95 分别达到 95.2%、94.5%、94.8%、96.7% 和 95.8%。模型包含 1.7 M 参数，计算量为 4.8 GFLOPs，模型大小为 3.8 MB，表明其具有轻量化鸡蛋分拣部署潜力。