Classification of Microscopic Images of Bacteria Using Deep Convolutional Neural Network

Classification of bacteria is essential in medical science for diagnosis of numerous diseases, treatment of infection, and trace-back of disease outbreaks. But it takes long time and huge human-effort to manually identify and classify a bacteria. With the advancement of technology, now the task of recognizing images from digital electron microscopes is being performed by computers based on machine-learning and computer-vision technologies. Besides, the latest generation of convolutional neural networks (CNN) have achieved impressive results in the field of image classification recently. Thus in this paper, we have investigated an approach to automate the process of bacteria recognition and classification with the use of deep convolutional neural network (DCNN). We have used the `transfer learning' method to retrain the `Inception DCNN model' with of a dataset of more than 500 microscopic images of 5 different bacteria species that are harmful to human-health. 20% images of the dataset were randomly chosen and separated, which were used to test the classification accuracy of the network. The retrained model was able to recognize and classify all 5 different species of bacteria, while the experimental results of prediction achieved accuracy of around 95%.