An IOT Platform for Disease Detection in Agriculture Using Convolutional Neutral Networks

Indian agriculture is at the nascent stages of adoption of Information and Commu- nication Technology (ICT) techniques for managing and improving farm output. ICT can prove to be beneficial for all farmers including small landholders, the most vulnerable to crop losses. With more than 40% of Indian population employed (di- rectly or indirectly) in it, agriculture is the most important part of Indian economy and crucial for India's growth. ICT techniques can enable the vulnerable farmers, especially small stakeholders, to take appropriate preventive / mitigative actions in case of crop diseases, adverse weather or even soil health. Machine Learning (and it's subset deep learning) and Artificial Intelligence have contributed to an explosive increase in application of computer science to complex science problems previously thought to be out of reach. Machine Learn- ing is the basis of this thesis, in particular Deep Learning and their subset, Con- volutional Neural Networks (CNNs). This section describes Machine Learning fundamentals and operation of Convolutional Neural Networks. Problems easy for humans such as game play or object recognition that were either difficult to describe mathematically or computationally too expensive have been able to uti- lize deep learning methods with great success. Especially, image recognition has seen a paradigm shift and use cases are popping up everywhere. Machine Learn- ing allows applications to predict more precise and accurate outcomes without being explicitly programmed.