Design and Development of Efficient Techniques for Leaf Disease Detection using Deep Convolutional Neural Networks

With the increase in the spread of crop diseases, there is a need to prevent and control its contamination so as to increase productivity and yield for the farmers. Plant Diseases have a detrimental effect on plants and animals and impact on market access and agricultural production. The proposed work use tomato leaf images for disease classification as tomato is one of the most important vegetable plants in the world and hence early detection of tomato leaf disease is required. Diseases of tomato plant include Bacterial leaf Spot, Yellow Curved, Late Blight, Tomato Mosaic and Septorial Leaf Spot. The dataset is taken online from plant village project. The idea of this paper is to take a dataset of the tomato leaf images with different leaf diseases and train it on a best model Convolutional Neural Network (CNN) and then use the obtained weights from the CNN for testing new tomato leaf images. The hybrid approach VGG16 with attention model is taken to achieve the best weights possible for testing and validation in the proposed model. The model showed the accuracy of 95.90 percent with hybrid approach. Performance analysis is done to identify the best model with good accuracy and also overcome the problem of overfitting.