anshvsingh/pneumonia-detection

🫁 Pneumonia Detection from Chest X-Ray Images

This model detects pneumonia in chest X-ray images using transfer learning with a ResNet architecture, trained on the Kaggle Chest X-Ray Images (Pneumonia) dataset.

⚠️ Disclaimer: This model is intended for educational and research purposes only. It is not a substitute for professional medical diagnosis. Always consult a qualified healthcare professional.

---

📌 Model Summary

Property	Details
Task	Binary Image Classification
Classes	Normal, Pneumonia
Architecture	ResNet (Transfer Learning)
Framework	TensorFlow / Keras
Input Size	180 × 180 × 3 (RGB)
Output	Softmax probability over 2 classes
Model File	xray_model.hdf5

---

📂 Dataset

The model was trained on the Chest X-Ray Images (Pneumonia) dataset available on Kaggle.

• Source: Kaggle - Chest X-Ray Images (Pneumonia)
• Total Images: 5,863 JPEG chest X-rays
• Classes: Normal, Pneumonia (Bacterial + Viral)
• Split:
  • Train: 5,216 images
  • Validation: 16 images
  • Test: 624 images
• Patient Demographics: Pediatric patients aged 1–5 years from Guangzhou Women and Children's Medical Center, Guangzhou

Citation:

Kermany, Daniel; Zhang, Kang; Goldbaum, Michael (2018), "Labeled Optical Coherence Tomography (OCT) and Chest X-Ray Images for Classification", Mendeley Data, V2, doi: 10.17632/rscbjbr9sj.2

---

📊 Metrics

Metric	Score
Accuracy	91.50%
Precision	91.00%
Recall	92.00%
F1 Score	91.00%

💡 For medical imaging tasks, Recall (Sensitivity) is the most critical metric — a missed pneumonia case (false negative) is more dangerous than a false alarm.

---

🚀 How to Load & Use the Model

Install dependencies

pip install tensorflow huggingface_hub pillow opencv-python numpy

Load the model

import tensorflow as tf
from huggingface_hub import hf_hub_download

model_path = hf_hub_download(
    repo_id="anshvsingh/pneumonia-detection",
    filename="xray_model.hdf5"
)
model = tf.keras.models.load_model(model_path)

Run a prediction

import numpy as np
import cv2
from PIL import Image, ImageOps

class_names = ["Normal", "Pneumonia"]

def predict(image_path):
    image = Image.open(image_path).convert("RGB")
    image = ImageOps.fit(image, (180, 180), Image.LANCZOS)
    image = np.asarray(image)
    image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
    image = image[np.newaxis, ...]

    predictions = model.predict(image)
    score = tf.nn.softmax(predictions[0])

    print("Prediction : {}".format(class_names[np.argmax(score)]))
    print("Confidence : {:.2f}%".format(100 * np.max(score)))

predict("chest_xray.jpeg")

---

🌐 Web Application

A live Streamlit web app is available for easy inference without writing any code.

👉 GitHub Repository: anshvsingh/pneumonia-detection

# Run locally
git clone https://github.com/anshvsingh/pneumonia-detection
cd pneumonia-detection
pip install -r requirements.txt
streamlit run xray_web.py

---

🔬 Approach

• Transfer Learning was applied using a pre-trained ResNet architecture
• The model was fine-tuned on chest X-ray images for binary classification
• Training was done on Google Colab using GPU acceleration
• TensorBoard was used to monitor training and evaluation metrics

---

👤 Author

Ansh V Singh

• GitHub: @anshvsingh
• Hugging Face: @anshvsingh

---

📄 License

This project is licensed under the MIT License — feel free to use and build upon it with attribution.