Last summer, I was in a conversation with my (Digvijay) grandpa and he mentioned the difficulties faced by farmers in diagnosis of crop diseases. Every year, they suffer immense loss and this happens solely because not every individual is experienced enough to predict plant diseases and figure out proper medication.

What it does

We built an application which helps in crop disease diagnosis. Farmers can just take a picture and upload it to the web app and it will predict whether the plant is suffering from any disease. This can be developed further to add support for varied range of plants, diseases and environmental & regional conditions helping perform tasks faster and more accurately.

How we built it

We leveraged computer vision, a subset of AI, to perform image processing. The model is trained onto 2000 images of cotton leaves with/without disease. The model was trained in cloud and used tensorflow & keras.

Challenges we ran into

It was difficult to find the dataset & the computational power along with time. We eventually ended up using PlantVillage & other open source datasets from Kaggle.

Accomplishments that we're proud of

We are glad that we were able to examine the ground reality of the situation and bring this major issue into notice. We built a web app using the model which acts as a proof-of-concept.

What we learned

It was amazing talking to farmers and conducting interviews. We learnt about the administration of Indian villages. Technically, it was a great journey of finding datasets, training the model and using it in our app.

What's next for Bean

We will be launching the app for public in the next 2 months. We have other technical improvements in our roadmap including better accuracy, more range of plants, faster results. Along with that, we are adding accessible guides for medication of the diseases.

Built With

Share this project: