A big issue in agriculture is the wastage of crops and food due to disease. Often, by the time the farmers realize there is disease present, it is far too late. Our apps hopes to tackle this problem, by providing a quick and easy way for farmers to quickly determine the health of their crops, allowing them to act accordingly and in a much more timely manner.
What it does
It leverages the power of Convolutional Neural Networks to classify images by first being trained on a dataset of images which create a mathematical mapping of every pixel value to a classification output.
How we built it
The neural network was created in Python using Tensorflow, and the output of this model can be linked via the TensorflowLite API for deployment to mobile applications. We created an app for both Android and iOS.
Challenges we ran into
Too many to count, but here's a few that stand out:
- Trying to standardize the front-end UI design for android devices to try and conform to as many screen layouts as possible due to the wide variety of android phones available
- Integrating the tensorflow model into Java for Android deployment
- Data collection, there was no good dataset available for our purposes, and thus we had to manually create one
What we learned
- Android development workflow
What's next for CropIT-Android
- Train a better and more diverse model
- Integrate the model with android