Using deep learning and AI to tackle global food insecurity and undernutrition

Building on the ongoing EI-Syngenta collaboration between the Zhou lab at EI and the Benjamins team at Syngenta, we are utilising the machine-learning based SeedGerm system to acquire high-resolution and high-frequency crop seed growth image series for automatically scoring key traits such as germination rate and germination timing. This PhD project will extend this study to use more complex and robust deep learning algorithms to improve the analysis framework, so that we can (1) develop a cloud-based training platform for academic and industrial users to label image datasets for crop species (i.e. corn, pepper and wheat); (2) build an open software system to integrate different learning architectures to perform sound phenotypic analyses; and, (3) establish neural network architectures to predict seed quality using phenotypic traits under varied experimental conditions (e.g. chemical treatments, temperature, and humidity), focusing on early establishment and reproductive (i.e. flowering) stages. Research outputs and machine learning toolkits produced in this project will be applied to Syngenta's ongoing seed improvement programme as well as the phenotyping tasks led by the Zhou lab in BBSRC's Designing Future Wheat.