Plant Disease Detection and Classification with Deep Learning

Deep Learning has gained massive attention in the academia for their trans-formative ability in creating a self- learning AI process in the fields of computer vision, trend prediction and natural-language-processing by utilizing neural networks which have been inspired by observing the workings of the human brain. Developed nations witness a trend shift where populations employed in agriculture drops from 60% to under 25% because of modernization. India is currently observing this trend of reduced agricultural participation from its population, a natural occurrence from when a country moves from a developing to a developed state. According to a report from the Ministry of Agriculture & Farmers Welfare. Since 2001, over 7.7 million farmers have quit farming and have shifted to urban occupations. From the latest consensus data on average, over 2,000 farmers quit the status of “Main Cultivator� every single day in India. This project aims to create a convolutional neural network capable of identifying plant types and the state of the plant i.e. its accompanying affliction if any which may negatively impact its yield, utilizing the recent and effective techniques acknowledged and published in the academia as a mean to assist the inexperienced workforce in an attempt to upend the lack of generational farming knowledge passed down through the centuries between farmers, reduce workforce training time and prevent catastrophic agricultural and financial failures. Utilizing dense net connections, convolution, and pooling; this project has created and trained a neural network model for over 50 epoch cycles until satisfactory to reach over 90% levels of accuracy with validation data.