Real-time optical acquisition and classification system for microbiology applications

This paper presents an image-based real-time microorganism classification system using deep learning object detection methods, namely YOLOv5. This prototype was developed using an optical kit consisting of a high amplification optical tube, as well as high sensitivity and spectral domain CMOS sensors, coupled to a dedicated deep learning application, a Nvidia Jetson Nano computer, running the trained state-of-the-art YOLOv5 architecture. For this purpose, an image dataset was assembled using 1193 images of the Escherichia coli (E. coli) bacteria taken from traditional microscopy and a digital camera. Each image was thoroughly labelled. Image data augmentation was also used to extend data. The classification system showed promising results, indicating that there is potential for further development of the model and setup. The proposed prototype lays a solid foundation for improving classification results in portable applications without the need for traditional but time consuming laboratory preparations for the analysis and identification of microorganisms.