Plant disease detection using deep learning based Mobile application.

Crop disease serves as a major threat to the farming sector. Due to the increased utilization of smartphones, it is now possible to leverage the technology and apply it for the betterment of the farming sector. The agricultural sector struggles in supporting the ever-growing global population, moreover, plant disease reduces the amount of food production and quality of the food. Losses may be a cataclysm, but on an average, it affects almost 45% of the production of major crops. Farmers often spend a lots of money on disease management of the crops and each crop is vulnerable to a particular disease that affects the quality and final yield. But, lack of proper technology, results in poor disease management, soil pollution, and the outcome may be devastating. In addition, plant diseases also affect the food chain supply, destroy the natural ecosystem and contribute to exacerbating environmental issues. These problems can be eradicated by adopting deep learning algorithms to analyze and visualize the current condition of the crops. With application built using deep learning, it is now possible to accurately detect crop diseases thereby reducing the effects of crop disease on food supply. Thus, correct disease detection followed by the management of identified diseases, thereby increasing food production and maintaining the quality of the food is achieved by deep neural networks. The proposed model uses MobileNet architecture along with complex hidden layers fine-tuned with Keras tuner on the dataset containing 12,318 images. We proposed an enhanced MobileNet scalable model with better generalization on large sized unified dataset constructed from various smaller sized dataset for better features' extraction and representation. The proposed model classifies the input in 64 different classes for 22 different sets of crops and achieved an accuracy of 95.94%. Further, the model is inculcated with our Android application – Plantscape for a better user experience fusioned with serene user interactions. [ABSTRACT FROM AUTHOR]