In this repository I have utilised Convolutional Neural Networks to predict whether the images detect a covid+ve patient, viral pneumonic patient and a normal patient.
The dataset is Covid 19 Radiography Database which has been updated as on 5 January 2021. The dataset has been imported from Kaggle with the following link:-
https://www.kaggle.com/tawsifurrahman/covid19-radiography-database/download
I have used two methodologies for this classification.
Krish Naik Youtube Video on Keras Tuner:- https://www.youtube.com/watch?v=Lx16T9cl5ng
1) CNN Using Keras Tuner
I used Keras tuner for hyperparamter tuning and created a very simple convolutional neural network having only two convolutional layers and two dense layers.
It just took around 2 minutes to train and acheived accuracies of around 97% in training set and 95% in validation and test sets and acheived excellent results
in a very minimal time.
2) Transfer Learning Using VGG16 Model
So here I implemented the classification using VGG16 Model. One layer was added after it so as to get our output and the pretrained model wasn't trained.
It took around 27 minutes to run for a single epoch and acheived a test set accuracy of around 93%. It also acheived excellent results on precison,recall
and f1 scores for each of the classes