Abstract

Agriculture plays a key role in India’s economic sector. More than 75% of the world’s population is dependent on agriculture, with most of its GDP coming from agriculture. Climatic and other environmental changes have become a major threat to agriculture. Grapes are a well-known fruit crop in India and are considered very important from a commercial point of view. However, there is a loss of 10-30% in grapes due to diseases. Grape diseases can cause significant losses to farmers and their grape production if not detected and treated early. Downy mildew, powdery mildew, leaf blight, Esca and black rot are the major grape diseases. Machine learning is a very effective solution to solve this problem. According to our research, convolutional neural network (CNN) is the most popular deep learning algorithm widely used in plant disease detection. In this paper, we did comparative analysis between CNN and AlexNet architecture to detect the diseases in grape plant and compared the accuracy and efficiency between these architectures. We used CNN algorithm and achieved an accuracy of 95.84% and AlexNet is a kind of CNN architecture used and achieved an excellent accuracy of 98.03%. The final result shows that the AlexNet architecture obtained higher accuracy than the CNN algorithm. In this work, an Android application has been designed to detect grape disease. When a farmer captures or uploads a photo of a diseased grape leaf, the mobile app predicts the disease and offers solutions to reduce the risk of the disease.