Diseases in monoculture farms can spread easily and significantly impact food security and farmers' lives. We aim to create a solution that uses computer vision for the early detection and mitigation of these diseases.

What it does

Our project is a proof-of-concept for detecting plant diseases using leaf images. We have a raspberry pi with a camera that takes an image of the plant, processes it, and sends an image to our API, which uses a neural network to detect signs of disease in that image. Our end goal is to incorporate this technology onto a drone-based system that can automatically detect crop diseases and alert farmers of potential outbreaks.

How we built it

The first layer of our implementation is a raspberry pi that connects to a camera to capture leaf images. The second layer is our neural network, which the raspberry pi accesses through an API deployed on Digital Ocean.

Challenges we ran into

The first hurdle in our journey was training the neural network for disease detection. We overcame this with FastAI and using transfer learning to build our network on top of ResNet, a complicated and performant CNN. The second part of our challenge was interfacing our software with our hardware, which ranged from creating and deploying APIs to figuring out specific Arduino wirings.

Accomplishments that we're proud of

We're proud of creating a working POC of a complicated idea that has the potential to make an actual impact on people's lives.

What we learned

We learned about a lot of aspects of building and deploying technology, ranging from MLOps to electronics. Specifically, we explored Computer Vision, Backend Development, Deployment, and Microcontrollers (and all the things that come between).

What's next for Plant Disease Analysis

The next stage is to incorporate our technology with drones to automate the process of image capture and processing. We aim to create a technology that can help farmers prevent disease outbreaks and push the world into a more sustainable direction.

Share this project: