Deep Transfer Learning for Enhanced Blackgram Disease Detection: A Transfer Learning - Driven Approach

In the field of agriculture, where various crop growth enhancement techniques are required. One such crop “Blackgram” is considered to be having various health benefits, due to which its growth and disease diagnosis becomes an important part. Out of the various techniques widely used one of the techniques for detection and classification of diseases includes Deep Learning models. This paper intends to make use of deep learning techniques such as Resnet152V2, VGG19, and InceptionV3 models which prove to provide the best results on 1007 images of Black gram Plant Leaf Disease (BPLD) dataset. The diseases that are classified are Anthracnose, leaf Crinckle, powdery Mildew, Yellow Mosiac and Healthy leaves. Performance metrics used in this paper are Training and Testing accuracy, Precision, recall, and F1 score. The study proves to provide the best results with the highest training accuracy of 98.80% and testing accuracy of 94.74% with the InceptionV3 models compared to ResNet152V2 and VGG19.