Hi, my name is Andy Barrett-Lennard, currently a student enrolled in a Masters of IT specialising in Software Development and Data Science. As part of our last semester of the degree we all have to undertake ITC571 - Emerging Technologies and Innovation. This is the capstone unit of the degree where we will make our contribution as part of our masters. As part of this process we need to document our project as we go along, this series on my blog will serve this purpose.

A bit about me

I am a trained winemaker with a BSc in Viticulture and Oenology. I have been working as a winemaker for a little over 12 years now, with the last 7 being the production manager at a relatively prestigious winery in Margaret River. However I have always said if I wasn’t a winemaker i’d have a career in IT.

My hobby has always been IT, in all forms, though with a large focus on system administration. I have always dabbled in tech and as the other entries in this blog show, enjoy self-hosting and making things as challenging as possible for myself! The masters was a way for me to formalise this hobby as well as broadening my knowledge base into coding and data science, two things I have always wanted to know how to do.

The project domain

Thus my two passions combined in the project idea for ITC571. We are currently under-taking a trial at work with a start-up that uses computer vision to assess bunch counts in the field and attempt to generate accurate yields estimates and spatial variability maps. There has been significant teething issues with this tech and as such I wanted to look at the current state-of-the-art in this domain.

However during my initial research for the problem I encountered two recent reviews of the field namely:

Mohimont, L., Alin, F., Rondeau, M., Gaveau, N., & Steffenel, L. A. (2022). Computer Vision and Deep Learning for Precision Viticulture. Agronomy, 12(10), Article 10. https://doi.org/10.3390/agronomy12102463

Ferro, M. V., & Catania, P. (2023). Technologies and Innovative Methods for Precision Viticulture: A Comprehensive Review. Horticulturae, 9(3), Article 3. https://doi.org/10.3390/horticulturae9030399

But one thing stood out to me while reading these two reviews, they focussed mainly on yield estimates and also spatial variability for things such as vigour. There seemed to be a lot of research and exciting development in the use of computer vision and analysis in pest and disease monitoring, but these reviews had not focussed on this particular aspect. Therefore I decided to focus on this problem set for my project topic.

There are a lot of interesting facets to this problem, namely around how results you can obtain in a lab translate to the messy domain of the field. From the dirty data to be interpreted by any algorithm to the limited compute and networking available on-farm. Hopefully this review can shed some light as to where we are on solving these hard problems.