Deep Leaf Disease Prediction Framework (DLDPF) with Transfer Learning for Automatic Leaf Disease Detection
T. Vijaykanth Reddy, K. Sashi Rekha
Abstract
Agriculture in India plays vital role in the economy and growth of the country. However, technology driven innovation in this field known as Precision Agriculture (PA) is still in its infancy. Nevertheless, there are significant improvements with technology innovations. With the emergence of deep learning as element of Artificial Intelligence (AI), it is made possible to bring about technology into agriculture activities. One of the activities that can be automated is leaf disease detection. With scalable access to cloud computing resources, this area of the research has attracted many academia and researchers. The existing approaches with Convolutional Neural Network (CNN) exhibited shortcomings in terms of adaptation and reuse of learned outcomes. This paper has used CNN with pre-trained deep models with transfer learning to fill this gap. A framework known as Deep Leaf Disease Prediction Framework (DLDPF) has been proposed by integrating CNN with AlexNet and GoogLeNet cascade inception. The underlying algorithm is known as Cascade Inception based Deep CNN with Transfer Learning (CIDCNN-TL). Keras and TensorFlow along with Python data science platform are used for implementation of the proposed framework. The proposed framework DLDPF is compared with many deep learning models such as AlexNet, GoogLeNet, VGGNet-16 and ResNet-20. Apple leafs datasets is use for empirical study. The experimental outcome exposed that the DLDPF outperforms the shape of the art deep learning model for automated prediction of lead diseases.