X-Ray Image Classifier Using ResNet-18 (PyTorch)

A medical imaging classifier built by fine-tuning a ResNet-18 backbone on X-ray data. Includes GPU-optimized training, augmentation pipelines, transfer learning, and evaluation scripts useful for healthcare AI prototyping.

---

🚀 Quickstart Demo (For Reviewers)

Run a single-image inference demo:

pip install -r requirements.txt
python demo_predict.py

This demonstrates preprocessing, feature extraction, and final classification.

---

📦 Full Training Run

Train the full ResNet-18 classifier:

python train_resnet18_xray.py

Includes:

• Frozen → unfrozen training phases
• Augmentations (resize, normalize, flips)
• ImageFolder dataset loading
• Accuracy reporting
• Transfer learning best practices

---

📁 Files

train_resnet18_xray.py   # Full transfer-learning training script
demo_predict.py              # Example inference on a sample X-ray
requirements.txt         # Dependencies

---

🏗 Overview

The pipeline uses:

• Pretrained ResNet-18 backbone
• Custom classification head
• Cross-entropy loss
• Optional mixed precision
• Data augmentations for robustness

This repo demonstrates applied computer vision engineering for medical imaging tasks.

---

📂 Project Structure

.
├── train_resnet18_xray.py
├── run_demo.py
├── requirements.txt
├── CONTRIBUTING.md
└── SECURITY.md

---

🧱 Architecture Overview

At a high level, the training system looks like this:

Input images (H x W x 3)
        |
    torchvision.transforms
    - Resize to 224x224
    - Data augmentation
    - Normalization
        |
        v
  Pretrained ResNet-18 backbone
        |
        +--> Convolution + Residual Blocks
        |
        v
 Global Average Pooling
        |
        v
  Fully-connected classification head
  (replaced with num_classes = 2)
        |
        v
  Softmax (via CrossEntropyLoss)

---

🤝 Contributing

See CONTRIBUTING.md for contribution guidelines.

---

📄 License

MIT License. See LICENSE for details.