Leafy

Inspiration

Our major source of motivation for leafy was an incident that took place recently. My mother and her friends are very enthusiastic about botany. My house has a big garden with innumerable plants. She usually spends a lot of her time catering to all their needs. Once my neighbour’s plant recently caught a infectious plant disease which was left undiagnosed. Due to this, the disease got communicated to our plants leading to their gradual death. Being nature lovers, it was a very traumatic experience for all of us. It made me wonder about the huge losses that the agriculture sector incurs because of the lack of right awareness at the right time. This prompted me to prepare a mobile application to prevent its reoccurrence.

What it does

The backend uses a well trained CNN based model which is trained by dataset comprising of over 30000 plant images. It can accurately predict and classify leaf diseases to a degree of 94%. If the model finds the detected disease to be infectious, it notifies all the nearby mobile phones(in radius < 2 km) to take precautions to prevent the disease.

How we built it

CNN model using Tensorflow, Scikit-learn, and OpenCV App using android studio API using flask(python) and around 5 cups of coffee ☕️

Challenges we ran into

We used our knowledge of java, python and html with a bit of css. Building a CNN model which could private the highest level of accuracy and F1 score was the main challenge for us. Another challenge was trying and testing the perfect combination of the CNN model parameters achieved desired results.

Accomplishments that we're proud of

Through the course of this hackathon, we learnt the use of Google cloud AI notebooks which will go a long way in helping us become better hackers. The fact that we were able to document our imagination and ideas into code successfully and complete this project in time is probably a big accomplishment for both of us.

What's next for Leafy

Just like we build Leafy for leaves, we can extend and develop another model which would be able to analyse soil through photos and predict about any lack of nutrients.

Room No:172 Team Name: MTC Hackers

Built With

• android-studio
• flask
• java
• opencv
• python
• tensorflow