Design And Implementation Of Bacterial Target Detection System Based On YOLOv5

In recent years,with the development of bioinformatics,deep learning techniques are becoming more and more common to solve the problem of cell image recognition.In microbiology,the growth period and characteristics of bacteria can be better understood by counting and morphological classification of bacteria.With the continuous development of instruments and equipment,a large number of microbial image information is produced rapidly,but manual data analysis is time-consuming and laborious,which is far from meeting the needs.Because the bacterial target is small and the image usually has some noise,the accuracy of common target detection methods is limited.In addition,it is difficult for biological researchers to use complex computer methods to analyze cell images.Therefore,a set of perfect graphical interface system can reduce learning costs,improve statistical efficiency,has universal significance.Based on the above background,this paper carried out the following research and design work: First,Bacillus subtilis was cultured in the biological laboratory,and the confocal microscope image and flow cytometer image data were used as data sources to make the corresponding data set.Then,YOLOv5 method was used as the basic framework,and an improved T-YOLOV5 method for bacillus subtilis detection based on attention mechanism was proposed.Aiming at YOLOv5's low accuracy on overlapping targets and small targets,CBAM attention module and SCA loss function are introduced to improve the mean average accuracy about.Aiming at the problem of too long training time,Ghost Net related module was introduced to shorten the model training time about while maintaining the original accuracy almost unchanged.In this paper,a web-based Bacillus subtilis detection system was designed based on this model,which provided cell detection,classification statistics and recording functions for laboratory researchers.Finally,this paper designs the corresponding Docker and Dockerfile files for this system,providing support for further development and optimization of this system.