- Keras
- numpy
- openCV
- Tkinter
About
Here you can see two chest X-rays, first being normal and second showing signs of pneumonia. You can clearly see that the right has areas of "haze", alluding to presence of pneumonia.
How could we get software to identify which one has pneumonia then?
Well, that's exactly where Convolutional Neural Networks (CNN) comes into play.
CNNs are commonly used to classify images and this repository contains CNN that I created to solve that problem with an accuracy of 89.7%!
GUI
Using Tkinter I created a GUI which allows the user to see the model's prediction on the unseen test data.
Model
The model was created using Keras.
The best model best_model.h5 achieved 89.7% accuracy on the unseen testing data.
This was achieved by using Keras Tuning to optimize the hyperparameters and the number of layers. The resulting model looks like:
