We chose to follow prompt #2. We all know that corn is one of the most produced crops in the world. The health of the crop is important for all farmers as it affects their harvest and their revenues. We decided to use artificial intelligence to diagnose corn crop disease in real time.
What it does
In an iOS app, the farmer can take a picture of his corn crop and the app will diagnose whether the crop is healthy or not. If it isn't, it will diagnose the disease it has.
How we built it
The development was a two-stage process. First was the part where we worked with big data. We mined hundreds of images of diseased and healthy corn to eventually train a convolutional neural network. We then deployed the model into a tflite file that is compatible with mobile platforms. We integrated this model in our iOS app that does the work for the farmer.
Challenges we ran into
-Finding a lot of data: As sad as it seems, we weren't able to find many pictures of diseased corn. This may have impacted the accuracy of our results, but our model does work otherwise (70% accuracy).
-Front End Issues: We took a lot of time debugging to make the camera module work with the classification engine.
Accomplishments that we're proud of
-Despite the small dataset, we had a relatively high accuracy (70%).
-We finished what we wanted to accomplish with the time given.
-This system is modular where you can expand the categories of classification (e.g. identify more crops!)
What we learned
-Training a model is the easy part, getting the data is the hardest part.
What's next for Cornado
-Expand the categories of classification, gather more data, and train our model to be more accurate.