This repository contains code and report for the COVID19 Chest XRay image classifications by using Deep Learning models. The problem is multilable classification problem in nature and hence we used Transfer Learning technique along with Focal loss as critarion. It is a multi label classification problem in nature. The dataset contains 3 classes ['COVID19', 'NORMAL', 'PNEUMONIA']. The dataset is available here
https://drive.google.com/drive/folders/1-_XiDnOEloIdIUN7bpJbrUPalKv8Zn0m?usp=sharing
Data distribution is as follows:
Batch size is 8 a batch example is shown as:
We used pretrained VGG16 model and replace its fully connected layers. The FC layer architecture is as follows.
Freezes all other layers except these FC layers and train this model on our dataset. The exparimental steps are as follows: momentum = 0.9 learning rate = 0.001 optimizer = Stochastic Gradient Descent critatrion = Focal Loss
and trained the model. The trained model can be found here:
https://drive.google.com/file/d/1X3plx3WiYD1R8CsDr-s2V7LXdMXFZve_/view?usp=sharing
The loss curve is:
The accuracy curve is:
The confusion matrices are:
F1 score is:
