Advancements in Plant Leaf Disease Identification Using Deep Learning and Machine Learning Perspective

Agriculture is the major occupation of people in India. Around 70% of actual population depends on agriculture. Potato is one of the major cash crops in agriculture industry and is used on a huge scale to produce clothes. Every year, many plants suffer damage due to diseases and infections, leading to the wastage of valuable resources. This research proposes a plant disease prediction system built on a deep learning and machine learning methodology in an effort to lessen this loss. This system will predict whether the plant or leaf is healthy or diseased and if the leaf or plant is diseased then it will suggest proper precautions and measures to prevent and cure the disease. By this the agricultural industry will get benefited and our country will move towards digital agricultural era. The method proposed in this paper makes use of deep learning’s power for feature extraction and machine learning classifiers for classification purposes. Convolutional neural networks are used to detect features, with a validation accuracy of 98.97%. These features are then fed to classifiers such as K-NN, SVM, Random Forest, MLP, and Naive Bayes. With an accuracy of 99.07%, the K-Nearest Neighbors classifier demonstrated remarkable performance. Even with its minor decrease, the Random Forest classifier managed to achieve an impressive accuracy of 98.30%. The SVM and MLP classifiers also demonstrated their capacity to accurately classify plant leaf diseases, with outstanding accuracy rates of 97.89% and 98.47%, respectively. The Naive Bayes classifier also proved to be effective, scoring an impressive 97.10% accuracy.