Using machine learning to identify and diagnose crop disease

Crop diseases can cause major yield losses and the ability to detect and identify them in their early stages is important for disease control. Machine learning methods, in particular deep learning, have shown promise in classifying multiple diseases across many different crop types. In this chapter we give an introduction to how deep learning for image analysis and classification works and explain the requirements for collecting a dataset of plant disease images for use with deep learning networks. We discuss the results and successes of various previous studies and highlight pitfalls with individual methods. It is clear that deep learning is capable of handling complex disease classification problems where one disease is present. There is plenty of room for growth to work with the presence of multiple diseases in a single image or to quantify the amount of disease present.

Table of contents
1 Introduction
2 A quick introduction to deep learning
3 Preparation of data for deep learning experiments
4 Crop disease classification
5 Different visualisation techniques
6 Hyperspectral imaging for early disease detection
7 Case study: identification and classification of diseases on wheat
8 Conclusion and future trends
9 Where to look for more information
10 References