Detection of Pepper and Tomato leaf diseases using deep learning techniques

Agriculture industry assumes a big role in several countries' economies by offering a few advantages including food, national income, and employment opportunities. However, it confronts a major challenge which is plant disease. Around 42% of the world's total crops are destroyed yearly by diseases. So, the identification of plant disease in an earlier stage is critical for minimizing the use of pesticides and reducing crop loss. This research is focusing on the recognition and classification of various tomato and pepper diseases based on Deep learning techniques. Two Convolutional neural networks (CNN) models Resnet 152 and Resnet 50 are used with and without data augmentation. This technique is used when the datasets are small, it expands the number of training images without adding new images. A self-dataset of 488 images of tomato and pepper diseases leaves are collected in Monastir, Tunisia region under the supervision of an agricultural engineer working in the field of plant protection in the Minister of Agriculture. The trained model with data augmentation gives better results than without applying it.