ST00L: Volumetric Stool Analysis

Inspiration

It started as a joke. Until we realized that this could have potential to do good in the world. In third world countries, sanitation is not as developed and so digestive tract diseases are more prevalent. As access to a doctor can be difficult, we wish for our app to help analyze stool to see whether there is a concerning issue.

What it does

The app allows you to take an image of stool and will give you an approximate size of the stool using a convolutional neural network.

How we built it

We created our training dataset using blender to generate images of stool in toilets. We then created a convulutional neural network using keras and trained it with this dataset. We then saved our neural network and incorporated it into the android app that we created using android-studio.

Challenges we ran into

We struggled with learning how to create an app since none of us have ever programming an app before. Luckily, there were many helpful tutorials online that explained how to use android-studio.

What we learned

Creating a convolutional neural network is very difficult and requires lots of time and patience.

What's next for ST00L: Volumetric Stool Analysis

We wish to expand our training data set to allow for a better convolutional neural network to be made (with greater accuracy results). We wish to include more features for a user to include (along with the image) to allow us to better analyze the stool and look for symptoms and give the user more information such as