Paper Title
Recent Advances in Image-Based Biosensor for On-Site Detection of Agricultural Infections Using Deep Learning

Abstract
Image-Based Biosensor technology has given new heights to plant diseases diagnosis by providing non-destructive, cost-effective, user friendly and onsite early detection system. Rhizome rot is one of the most dreadful infections of folk medicinal plants resulting in abrupt declination in both quality and yield sometime loss of entire plant. It is crucial to reduce the losses by catching the pathogens at an early stage and to prevent the spread of infection. The present review describes the conventional detection methods and recent advances in rhizome rot detection importantly on artificial intelligence based image biosensors. Conventional detection methods are the “gold standard” in rot detection which are mainly based on symptoms, fungal colony based, morphological observation and biochemical identifications. In recent years the functioning of deep learning in image recognition and classifications has made remarkable progress. The present study aimed at developing sensor by combining effective deep learning models, trained with images of infected leaves and rhizomes, for rapid detection of rhizome rot. Images of rhizome rot infections in various infective stages along with the bands obtained in the immunochromatographic based nano sensors basing on the intensity of the (band) infection from mild - acute - chronic - severe – very severe infections by comparing with healthy plants can be used. Using dataset of above mentioned images of infected and healthy plant samples, a deep convolutional neural network can be trained to identify the initial rhizome rot infections or its absence thereof. The trained model designed to achieve an accuracy of 99% can be considered for the feasibility of this approach. In addition, we also described the methods using trained model to treat the infection with the recommended dose of fungicide which will be available in the biosensor. This approach of training deep learning models will help to design smartphone-assisted biosensor for diagnosis of infections in rhizome containing folk medicinal plants like turmeric, and ginger on a larger macro scale. It will help in developing different types of disease control strategy and improve the harvest quality and accuracy in agriculture. Keywords - Rhizome Rot; Deep Learnings; Image Processing; Digital Nano-biosensor