According to the latest Statistics Canada report, 2018 registered the most significant decrease in net farm incomes in Canada at 45.1%, the largest percentage decrease since 2006, the year records started. This followed a 2.8% decline in 2017. The loss in income is mainly attributed to rising operational costs, primarily fertilizers and pesticides, and relatively flat revenues from the sale of agricultural crops and livestock, amidst changes in mandated regulations to limit carbon and nitrogen oxide based emissions in agriculture. All of this pre-COVID. In 2020, tourism was also stopped, further affecting farmers' chances of catching up on loans and payments. We wanted to use technology that would help farmers not only for their tourism attraction struggles, but also to address as much as possible the increasing costs on farmers.
What it does
For this hackathon project, we built an app (smartphone, web app) that farmers can use to manually monitor their harvests' health. All the farmer needs to do is to take an image with their app. The app uses machine learning to help farmers promptly detect diseases in grapes, allowing them to effectively know which bacteria and fungi are present as well as providing insight into [potassium levels] soil nutrient concentrations, potentially allowing them to address fertilizer and pesticides issues sufficiently and effectively. This data-driven approach reduces farmers' operational costs and ensures a more sustainable pathway into healthier crops and a better harvest. Sustainability and environmentally conscious is increasingly becoming important among communities, thus, an attractive quality that spikes tourists' attention.
In addition, the machine learning model can run on a raspberry pi, allowing a continuous monitoring of grape trees' health against pests and diseases.
How We built it
We used CustomVision to create an image classification model. We then exported the TFLite model to be used to classify images into either normal or containing bacteria or fungi in grape, grape leaves, and grape trees.
Web app: flask (python)
raspberry pi: TFLite on raspberry pi/ python
Android: android studio/java
Challenges We ran into
Finding enough images to create an accurate and trustfulness model. Our model needs a lot more iterations and images to perform better.
Accomplishments that we're proud of
We're proud that we were able to create this technological solution within a short period of time, collaborating with teammates from different universities and backgrounds.
More importantly It works!!! check out the demo.
What we learned
We learnt about some of the challenges facing farmers in Canada, and the importance of tourism in farming. Agritech and technology can help farmers significantly, in terms of lowering costs and helping them monitor their farm.
What's next for Team 18
We're all still students, working on getting more involved in sustainability. Some of us are passionate about agritech and have been working towards other type of agritech solutions, such as using technology to reduce overnitrification of soil and the overall GHG emissions. Regarding the app, we would like to improve our dashboard and connect our web app the Minister of Agriculture and Agri-Food which could alert other farmers regarding epidemic diseases.