COVID-19 CT Detection

Overview

Pretrained (imagenet) ResNet50 with 3 classification. Achieves 97.2% test accuracy. Use Grad-CAM for infected area segmentation.

Image Enhancement

Gamma transform and CLAHE

Datasets:

[1]Chowdhury [2]IEEE

Train-Validation Split

	COVID-19	Viral Pneumonia	Normal
Total	756	1361	1422
Train	600	1200	1200
Test	156	161	222

Train Accuracy

Confusion Matrix

Heatmap for Infected Area