Sundry Bacteria Contamination Identification of Lentinula Edodes Logs Based on Deep Learning Model

Lentinula edodes logs are susceptible to sundry bacteria contamination during the culture process, and manual identification of contaminated is difficult, untimely, inaccurate. Aiming solve this problem, paper proposes a method for based on deep learning model Ghost–YOLOv4. Firstly, data set log was constructed. Secondly, in view problems that YOLOv4 network parameters too large detection speeds videos slow, backbone feature extraction replaced with lightweight network, GhostNet, enhancement PANet Yolo Head modules were completed. The modification reduced number improved speed network. Finally, introduced migration pre-training model, which computational pressure overfitting further performance Ghost–YOLOv4 Not only did constructed ensure accuracy contamination, but it also had better results real-time performance, provides an effective solution deployment target embedded equipment sheds.