Pneumonia xray analyzer

Inspiration

Lots of people every day die from pneumonia, one of the biggest causes is simply not being able to detect it

What it does

with just one x-ray and this analyzer, you can find the starting of pneumonia

How we built it

we built it on a tensorflow framework, we trained a model of around 5000 images from a kaggle dataset and used that model to compute inputed images

Challenges we ran into

at some point along the way the model simple didn't seem like it was giving the right output, it took a decent bit of tweaking, and a lot of retraining but eventually we got it to have a high success rate

Accomplishments that we're proud of

simply the fact that with high accuracy its able to predict pneumonia on non edited images, and its held up

What we learned

a big part is that we didn't really know how to do almost any image classification, and we had to not only learn, but implement it into a finished product in just 2 days, and though that we learned a lot about neural networks, and image classification

What's next for Pneumonia xray analyzer

a big nex step would be the ability to highlight "suspicious" or important parts of images that contribute to the models prediction, such as opacity in areas, structures, and things like that.

Built With

• python
• streamlit
• tensorflow