Texture Features Retrieval using Convolutional Neural Networks for Multiclass Classification of Plant Diseases

Early leaf detection is essential to India's rapidly growing agricultural industry. Early plant leaf issues identification and the creation of forecasting techniques are crucial for the agriculture industry to fulfill its task of nourishing the entire population. Recently, analysis of images and automated vision algorithms have been used to images of pepper plant, tomato and potato plant leaves in order to detect illnesses in these plants. These approaches are used to obtain leaf image characteristics using either deep learning frameworks or specially developed methodologies. This article proposes a VGG19 model, based on transfer learning to address plant diseases. To train the model, the Plant Village dataset has been employed. There are 15 distinct leaf classifications in this dataset. The RMSprop optimizer has been employed to train the VGG19 model across 40 iterations with a batch size of 64 and a minimal learning rate of 0.0001. With a training accuracy of 97.85% and a testing accuracy of 96.98%, the model has performed effectively when categorizing images of the plant disease leaves. This study can be applied in the agricultural sector to identify plant illnesses early.