Abstract
One of the most significant areas of precision agriculture research is the identification of diseases using images of plant leaves. Traditional farming employs experts to physically analyse crops for host disease plants, which is a labour-intensive, time-consuming, and possibly error-prone job because it is done by manually. Large farms can use data-driven strategies, cutting-edge technologies, and enhanced training and standardization procedures such as machine learning to solve these misclassification issues. Automatic detection of plant types and diseases is essential to identify the symptoms of diseases as early as possible. It is crucial to develop an automatic system to help with this task. In this research, we introduced a Green leafy vegetables and disease identification and classification system using the simple CNN, MobileNet, Resnet 50, VGG-16, and Inception V3 models. To train these models, a dataset of 1044 images was used, which was collected from a multisource of agricultural published articles. Data augmentation is technique that increases the dataset for better accuracy when training the model. The accuracies of each model were compared to each other. The MobileNet achieved an accuracy of 72% for the classification of diseases and 94 % for the classification of vegetables, with a training time of 157 seconds. The simple CNN and VGG-16 achieved an accuracy of 83% and 99%, respectively, for classifying green leafy vegetables. The Resnet 50 and Inception V3 achieved an accuracy of 93% and 83%, respectively, for the classification of green leafy vegetables, and MobileNet achieved an accuracy of 94% for the classification of Green leaf vegetable diseases. The VGG-16 achieved the highest accuracy of 99%, the MobileNet achieved the highest accuracy of 92%, while the Resnet 50 and Inception V3 achieved 91.7% and 22% accuracy, respectively.