-
They posted an xray of a chest with pneumonia.
-
The home page
-
They are talking about it.
-
More themes and colors
-
some of the code
-
how we built the AI image classification model
-
more of the code from colab for the AI
-
at the bottom right you can see it has 93% accuracy but when we trained it for a higher number of epochs, it went up to 98%
-
this is how it takes uploaded pictures, and at the bottom you can see some of the pictures that were uploaded. it got all of them correct.
Inspiration
We were about issues they are having health problems, inspired by Covid, and how people might want to talk to people in their community but might not be able to talk to each other if they don't have each others' contacts or phone numbers. So they can just get the link to the community's forum from google and then see what people are saying, so more people will be connected with ease. This forum can really be used to talk about anything, but we wanted to focus it on health problems - specifically respiratory issues. They can even get predictions and find out if they have covid or pneumonia chest diseases by uploading a picture.
What it does
A community forum that allows people to share their health issues by making posts and comments, and they can upload images to receive predictions (with AI) so that they know if they have an illness. There are different themes/colors they can choose between too.
How we built it
Built the AI with tensorflow/keras/python in colab and the web app in codesandbox with html.
Challenges we ran into
Were trying to figure out how to upload more datasets into the AI so that it could do more than just predict respiratory illnesses. Were trying for so many hours to get the uploads from the website to go to colab and for the results to be published in the website. Took several hours to go through tutorials, even with instructions and extra help on how to use tensorflow, as we had so many issues building it. Most of the time was taken up by trying to fix the AI and connect it to the website.
Accomplishments that we are proud of
93-98% accuracy with the AI when people upload their pictures to it, the HTML allows uploads and the python script does too. People get results and will learn if their xrays show disease or good health. People can make posts and talk about any health issues in the community and can make comments. We used covid vs healthy lungs, but when images of other chest diseases were uploaded, it sorted those as Covid Detected lungs, so basically it predicts if a disease is present.
What we learned
We learned that colab is a complicated way to run this program and that it is not efficient, and it would have been better if we used a different app.
What's next
Not just predictions for pneumonia - we will be able to predict any disease, including bacteria in plants and fruit leaves so that farmers can learn what is wrong with their crops and can figure out how to fix it. We just need to upload more training/testing sets, as the AI is pretty good at learning how to make predictions and it can be trained on any datasets. The forum feature is also going to be improved and people will be able to make replies and comments to specific posts.
Log in or sign up for Devpost to join the conversation.