Food security in Africa under serious threat from plant diseases and with the already strained food supply the problem is likely to grow. Pest and disease induced crop losses have devasted farms across the world but thanks to advances in machine learning farmers can detect diseases early through the use of computer vision in the browser or on a mobile device

What it does

It detects plant diseases when a user uploads leaf images with the affected areas and it also shares videos of modern farming practices. There is also a curated list of 4 videos for Maize, potato, soybean and tomato farming which are some of the main crops grown in Africa.

How I built it

I used pytorch and fastai to train a model on the plant village crop dataset, then used flask to build a webhook for Messenger. I used the wikipedia library to search more information on the predicted classes.

Challenges I ran into

First with heroku as I wanted to also add a question answering model but the free tier is severely limited for 2 models so I had to do without it and then embedding a video on messenger I then settle for just the link.

Accomplishments that I'm proud of

It works, somewhat!

What I learned

The Messenger Platform and how to make a webhook with python and flask

What's next for Agrotorch

Add a conversational AI framework like, deeppavlov, rasa to make the bot smarter

Grow it into a useful channel for African farmers

Share this project: