• 

  Photo of the app

Inspiration

It has long been known that pesticides on food can cause problematic health outcomes for those who consume them. However, fewer people realize that the same rationale applies to the farmers who use them. As a team composed of agricultural enthusiasts, with some having direct ties to these communities, we know firsthand the impact of the use of these chemicals on farmers. Research from the National Institute of Health has also strongly linked several forms of cancer and neurological conditions to farmers due to their prolonged exposure to chemicals such as Chlorpyrifos (which was recently banned). Notably, increased risks have been identified for non-Hodgkin lymphoma, leukemia, and prostate cancer. With Nutra, we hope to mitigate this insidious part of the agriculture production pipeline by giving farmers a precise estimate of how much pesticides in tons they need for their crops, bringing more certainty to a process that many once considered an art form.

What it does and how we build it

Nutra is an iOS application built to empower farmers with precision agriculture to give them estimates of the amount of pesticides they need for a given crop yield through machine learning. With this information, farmers can also receive notifications on when it is best to distribute said chemicals. The last component of our app is a list of suggestions on how to further reduce pesticide waste and exposure, thus safeguarding personal health. We created the farmer-facing part of the application in Swift because personal phones are one of the few things farmers may bring into the fields. When the grower first opens the app, they will be presented with a quiz, which is the necessary information to start our predictions. Our model uses a random forest regressor to predict the optimal pesticide use given the farmer’s unique farming circumstances. Sensors throughout the farm collect current data on weather and soil conditions to generate precise recommendations for pesticide application using the machine learning model trained from historical pesticide usage data. Data from sensors are stored in real time using the Kintone microcontroller and Kintone web database for easy IOT integration

Challenges we ran into

At first we ran into the problem of having too much data and that was really hard to comprehend in the 24 hours hackathon. This caused us to think on our feet and focus on models that had the key factors needed in predicting the amount of chemicals needed to reach a target yield instead of aiming to account for every possible environmental factor. Well also had issues integrating the custom model into iOS.

Accomplishments that we're proud of

We are really proud that we were able to build an app in Swift since many of the developers on our team were unfamiliar with the language. In addition that, we are also proud being able to implement our python base machine learning model into our application.

What we learned

The factors relating to the extent of pesticide dissemination in the environment and waste. Through this hackathon and subsequent research into our problem, we found that even the smallest aspects such as temperature and wind can affect the amount of pesticides needed for a given job. As a team, we also learned how to fail fast and build back better.

What's next for Nutra - Be Aware, Handle with Care: Pesticide Safety First

In the end, we would love to provide better predictions for our farmers for more plant species. In addition to that, we believe that farmers would find it helpful to be able to get the predicted cost for the pesticides that they use since it will allow them to budget for the future. Lastly, we envision that Nutria will work better with a family of sensors that will allow farmers to get live predictions.

Built With

• flask
• kintone
• python
• swift