Covid CT Classification

This is the implementation of a convolutional Neural Network on COVIDx CT

About the Dataset

• The following dataset contains CT images from 194,922 CT slices from 3,745 patients and 201,103 CT slices from 4501 patients respectively.

• Classes are zero-indexed with Normal=0, Pneumonia=1, and COVID-19=2

• The COVIDx CT-2 dataset is released as a directory of images (2A_images) and associated label files ({train,val,test}_COVIDx_CT-2A.txt) indicating classes and bounding boxes for the body region.

Link to download the dataset: Here

Downloading the trained Model:

Here

Accuracy Scores

After 3 epochs

basic cnn

• Train Loss: 0.026 | Accuracy: 99.146
• Test Loss: 1.190 | Accuracy: 76.404

Resnet

• Train Loss: 0.024 | Accuracy: 99.229
• Test Loss: 1.056 | Accuracy: 79.012

VGG-16

• Train Loss: 13.014 | Accuracy: 70.982
• Test Loss: 12.807 | Accuracy: 58.678