DDCMR2: A Deep Detection and Classification Model with Resizing and Rescaling for Plant Disease

Identification of plant diseases is necessary to prevent declines in agricultural product yield and quantity. Different deep learning (DL) models were proposed by researchers for the classification of various plant diseases. Models that employ deep learning require vast amounts of data, increase the amount of training time needed and make it challenging to apply them in small systems and devices. A deep detection and classification model with resizing and rescaling (DDCMR2) for plant disease is proposed as a deep convolutional neural network (DCNN) in this research, 14 separate plant datasets totaling 38 classes were used to train and evaluate the proposed model with auto resizing and rescaling as a layer. On the PlantVillage dataset, the performance accuracy is 98.02% for training and 97.09% for validation. The proposed model uses cache and prefetch, making it significantly faster to train than other models. While training on the first epoch, the training time is longer for any low-level systems, however, training on the second epoch with the aid of cache and prefetch is acceptable and quicker.