🦠 Detecting COVID-19 with Chest X-Ray using PyTorch

This project demonstrates the use of deep learning for image classification on chest X-ray images to detect COVID-19, Viral Pneumonia, and Normal conditions using PyTorch and ResNet-18.

🧠 Guided by the Coursera project: Detecting COVID-19 with Chest X Ray using PyTorch
📦 Dataset from Kaggle: COVID-19 Radiography Dataset

---

📌 Project Overview

The goal of this project is to train a convolutional neural network to classify chest X-ray images into three categories:

• Normal
• Viral Pneumonia
• COVID-19

Using transfer learning with ResNet-18, I aim to achieve high accuracy on a held-out test set. The final model reaches over 95% accuracy after just one epoch.

---

🗂️ Sections

1. 🧾 Introduction

This project is a deep learning-based image classification task using PyTorch. The model is trained to identify whether a chest X-ray shows signs of:

• Normal lungs
• Viral Pneumonia
• COVID-19

---

2. 🗃️ Preparing and Splitting the Dataset

I used the COVID-19 Radiography Dataset from Kaggle. The dataset was reorganized to separate training and testing samples. For each class, 30 random images were moved to a test folder. `python source_dirs = ['NORMAL', 'Viral Pneumonia', 'COVID-19'] class_names = ['normal', 'viral', 'covid']

Create test folders and move images

3. 🧰 Creating Custom Dataset

A custom ChestXRayDataset class was built using PyTorch’s Dataset abstraction. It loads images dynamically from the labeled folders. class ChestXRayDataset(torch.utils.data.Dataset): def getitem(self, index): ...

4. 🖼️ Image Transformations

Image preprocessing and augmentation included resizing, normalization, and horizontal flipping (for training only).

train_transform = transforms.Compose([...]) test_transform = transforms.Compose([...])

5. 🧪 Prepare DataLoaders

The dataset was wrapped using PyTorch DataLoader for batch processing and shuffling.

dl_train = DataLoader(train_dataset, batch_size=6, shuffle=True) dl_test = DataLoader(test_dataset, batch_size=6, shuffle=True)

6. 🔍 Data Visualization

To evaluate prediction quality, a custom function displays images with true and predicted labels. Correct predictions are shown in green; incorrect ones in red.

def show_images(images, labels, preds): ...

7. 🧠 Creating the Model

Used ResNet-18 with pretrained weights. The final fully connected layer was modified to output 3 classes.

resnet18 = torchvision.models.resnet18(pretrained=True) resnet18.fc = torch.nn.Linear(512, 3) We used:

Loss Function: CrossEntropyLoss

Optimizer: Adam (learning rate = 3e-5)

8. 🚂 Training the Model

Training was done for 1 epoch with early stopping if accuracy exceeded 95%. Periodic evaluations on the test set were performed during training.

def train(epochs=1): ...

9. ✅ Final Results

📈 Validation Accuracy Over Time:

Step 0: 27.78%

Step 20: 84.44%

Step 40: 88.89%

Step 60: 95.56%

🧠 The model hit the accuracy target and training stopped early.

🧪 Technologies Used

• Python
• PyTorch
• torchvision
• matplotlib
• PIL
• NumPy
• ResNet-18 (pretrained)