Deep Learning Based Fruit Crop Disease Classification Using Plants Leaf Imagery

Agriculture plays an important role in the economy, providing nutrition, food, and livelihoods to billions of people in this world. Fruits are a vital component of a balanced diet, acting as a rich source of antioxidants, vitamins, minerals, and fibres. To ensure its proper supply to the market, early diagnosis and treatment of fruit crop disease can play a significant role. In order to make this process quicker and more accurate, Deep learning (DL) techniques can be of immense help. In the proposed study, the implementation of four DL models: Visual Geometry Group VGG16, VGG19, EfficientNet-B0, and MobileNet is done on a subset of the Plant Village database, constituting 7,233 leaf images of apples and grapes divided into 8 different classes. Evaluation is done through various metrics, namely Precision, Recall, F1-score and Specificity. All the DL models performed well by attaining accuracy above 95%. MobileNet performed best among all DL models by achieving an accuracy of 99.38%, while VGG16, VGG19, and EfficientNet-B0 achieved an accuracy of 96.75%, 98.69% and 98.76% respectively.