Detection of egg appreance based on Fasternet and YOLOv5 model
[Objective]Efficiently identify eggs with defects on their appearance in the automatic production process.[Methods]Designed a detection model based on fusing Fasternet module and YOLOv5s.The model used the Kmeans++algorithm to re-cluster the dataset and optimizeed the bounding box.The Bottleneck module in the C3 structure was replaced by the Fasternet Block module to reduce the parameters and improve the percision in the process of detection.The Soft-NMS,a non-maximum suppression was utilized to improve the detection of eggs with similar features.The CBAM attention mechanism was introduced to enhance the function of extracting important features.[Results]Compared with the YOLOv5 original model,the experiment results showed that the mAP@0.5 and mAP@0.5:0.95 respectively had increased by 3.2%and 5.2%,respectively.The amount of calculation and parameters was reduced by 19.6%and 16.9%,respectively.Compared with YOLOv7-tiny and YOLOv8 models,the improved model has significant advantages.[Conclusion]The experimental method can optimize the detection percision and reduces the parameters in the detection of egg'appreance,so as to achieve the purpose of identifying defected eggs in the automatic production.Efficiently identify eggs with defects on their appearance in the automatic production process.
detection of egg appearanceYOLOv5FasternetKmeans++CBAM
国家科技期刊平台
NSTL
万方数据
