Skip to content

This repository contains code and results for COVID-19 classification assignment by Deep Learning Spring 2020 course offered at Information Technology University, Lahore, Pakistan. This assignment is only for learning purposes and is not intended to be used for clinical purposes.

Notifications You must be signed in to change notification settings

SobanMahmood/MSCS16012_COVID19_DLSpring2020

Repository files navigation

Deep Learning COVID-19 Assignment

This repository contains code and results for COVID-19 classification assignment by Deep Learning Spring 2020 course offered at Information Technology University, Lahore, Pakistan. This assignment is only for learning purposes and is not intended to be used for clinical purposes.

Results

(Confusion matrix and accuracy)

Task 1: Transfer Learning while Freezing the Feature Layers

VGG-16:

Accuracy = 96% Confusion matrix =

Infected(Pred) Normal(Pred)
Infected(Actual) 556 49
Normal(Actual) 5 880

Resnet-18:

Accuracy = 97% Confusin matrix =

Infected(Pred) Normal(Pred)
Infected(Actual) 582 33
Normal(Actual) 10 875

Task 1: Transfer Learning while Freezing the Feature Layers

VGG-16:

Accuracy = 97% Confusin matrix =

Infected(Pred) Normal(Pred)
Infected(Actual) 582 33
Normal(Actual) 10 875

Resnet-18:

Accuracy = 96% Confusin matrix =

Infected(Pred) Normal(Pred)
Infected(Actual) 592 23
Normal(Actual) 30 855

Dataset

###Assignment Part 1: https://drive.google.com/file/d/1-HQQciKYfwAO3oH7ci6zhg45DduvkpnK/view ###Assignment Part 2: https://drive.google.com/file/d/1eytbwaLQBv12psV8I-aMkIli9N3bf8nO/view

Weights

https://github.com/SobanMahmood/MSCS16012_COVID19_DLSpring2020/tree/master/weights

About

This repository contains code and results for COVID-19 classification assignment by Deep Learning Spring 2020 course offered at Information Technology University, Lahore, Pakistan. This assignment is only for learning purposes and is not intended to be used for clinical purposes.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published