The production of medicinal cannabis in a controlled environment is a complex, multi-stage process which requires detailed and frequent crop monitoring to protect both plant health and the quality of the resulting product. These activities are presently done by operators in person, which makes them labour-intensive and costly in terms of time. Likewise, there are limits on what operators can reasonably measure during a production cycle.

​

The application of artificial intelligence for medicinal cannabis production offers a means to automate some of these labour-intensive activities and improve the range of data they can collect. Several parts of the process stand out as being possible to automate:

1.         Determining when a batch is ready to harvest

2.       Monitoring plant growth during the flowering phase

3.       Monitoring flower bud development

The student will be based at the University of Aberdeen, embedded in a machine learning group that includes several other students working on agri-food-related problems. In addition, the student will benefit from co-supervision in plant modelling areas from our university partners in the University of Lincoln’s Lincoln Institute of Agri-Food Technology. On a day-to-day basis, the student will be responsible for setting the experimental setup required to collect new data, as well as organise and pre-process historical and new data; acquire deep knowledge of multimodal deep learning systems and how they can be combined with crop models; and develop and evaluate deep learning pipelines that can operate in sparse data scenarios, that can generalise well, and be robust at inference stage.

 

The students will be embedded in a unique environment that includes two universities and an innovative industry partner, Waterside Pharmaceuticals. Extensive training will be provided depending on the student’s background with the anticipated skills development involving expertise in machine learning and deep understanding of crop modelling.