Detecting COVID-19 with Chest X-Ray using PyTorch

Train a model on a COVID-19 Radiography dataset from kaggle, comprising nearly 3000 Chest X-Ray scans categorized into Normal, Viral Pneumonia, and COVID-19 classes. A practical implementation of Convolutional Neural Networks (CNNs) and optimization techniques.

Learning Objectives

• Create custom Dataset and DataLoader in PyTorch
• Train a ResNet model for Image Classification

Skills Practiced

• Machine Learning
• Deep Learning
• Statistical Classification
• PyTorch
• Medical Imaging

Model Performance

After training the model for 30 epochs, the following performance metrics were achieved on the test set:

• Test Accuracy: 87.25%

Classification Report for Test Set

Class	Precision	Recall	F1-Score	Support
Normal	0.97	0.85	0.91	1019
Viral	0.70	0.96	0.81	134
COVID-19	0.74	0.89	0.81	361

• Overall Accuracy: 87.25%
• Weighted F1-Score: 0.88