A comprehensive deep learning-based system for detecting pneumonia from chest X-ray images using YOLOv8 classification model and a user-friendly Streamlit web application.

LIVE LINK - https://x-ray-classification-8432.streamlit.app/

DATA SET LINK - https://www.kaggle.com/datasets/paultimothymooney/chest-xray-pneumonia

• Overview
• Features
• Project Structure
• Installation
• Usage
• Model Training
• Streamlit Application
• API Documentation
• Contributing
• License

This project implements an automated pneumonia detection system using state-of-the-art deep learning techniques. The system can analyze chest X-ray images and classify them as either normal or showing signs of pneumonia, providing confidence scores and detailed probability breakdowns.

• Real-time Analysis: Instant classification of chest X-ray images
• High Accuracy: Trained on a large dataset of medical images
• User-Friendly Interface: Web-based application for easy access
• Sample Images: Built-in sample images for testing and demonstration
• Professional Results: Detailed probability breakdowns and confidence scores

• Interactive Upload: Drag-and-drop or click-to-upload interface
• Real-time Processing: Instant analysis with loading indicators
• Visual Results: Side-by-side display of uploaded image and results
• Sample Images: Downloadable sample images for testing
• Responsive Design: Works on desktop and mobile devices
• Professional UI: Medical-themed interface with clear instructions

• YOLOv8 Classification: State-of-the-art deep learning model
• Binary Classification: Normal vs. Pneumonia detection
• Confidence Scoring: Detailed probability breakdowns
• Pre-trained Weights: Ready-to-use trained model
• Custom Training: Full training pipeline included

• Prediction Labels: Clear normal/pneumonia classification
• Confidence Metrics: Percentage-based confidence scores
• Probability Breakdown: Detailed class-wise probabilities
• Visual Indicators: Color-coded results (green for normal, red for pneumonia)
• Medical Disclaimers: Professional healthcare warnings

• Python 3.8 or higher
• pip or conda package manager

git clone <repository-url>
cd Image-Processing

pip install -r requirements.txt

python -c "import streamlit, ultralytics, PIL; print('Installation successful!')"

1. Start the Application:

   streamlit run app.py

2. Access the Web Interface:

   • Open your browser and go to http://localhost:8501
   • The application will automatically load

3. Using the Application:

   • Upload an Image: Use the file uploader to select a chest X-ray image
   • Download Samples: If you don't have images, download the provided sample images
   • View Results: See the prediction results with confidence scores
   • Analyze Probabilities: Review the detailed probability breakdown

• JPEG (.jpg, .jpeg)
• PNG (.png)
• BMP (.bmp)
• TIFF (.tiff)

The model is trained on a comprehensive dataset of chest X-ray images:

• Normal Images: 6,084 images of healthy chest X-rays
• Pneumonia Images: 7,744 images showing signs of pneumonia
• Total Dataset: 13,828 high-quality medical images

1. Data Preparation: Images are organized into normal and suffering categories
2. Model Configuration: YOLOv8 classification model with custom parameters
3. Training Execution: Automated training with validation
4. Model Evaluation: Performance metrics and confusion matrix generation
5. Weight Saving: Best and latest model weights are saved

cd yolov8_custom_training
python main.py

• Accuracy: High classification accuracy on validation set
• Confusion Matrix: Visual representation of model performance
• Training Metrics: Detailed training and validation curves
• Model Weights: Optimized weights for production use

• Cached model loading for performance
• Automatic error handling for missing model files
• Support for both best and latest weights

• Automatic image format conversion
• Temporary file handling for YOLO compatibility
• Error handling for unsupported formats

• Two-column layout for image and results
• Color-coded prediction indicators
• Professional medical disclaimers
• Detailed probability breakdowns

• Real training images for testing
• Downloadable sample files
• Authentic medical image quality

• Responsive Design: Works on all device sizes
• Professional UI: Medical-themed interface
• Error Handling: Graceful error messages
• Loading States: User-friendly loading indicators
• File Validation: Automatic format checking

def predict_image(image):
    """
    Make prediction on uploaded image
    
    Args:
        image: PIL Image object
        
    Returns:
        dict: Prediction results with class names, probabilities, 
              predicted class, confidence, and class index
    """

@st.cache_resource
def load_model():
    """
    Load the trained YOLO model with caching
    
    Returns:
        YOLO: Loaded model object or None if error
    """

This application is for educational and demonstration purposes only. It should not be used for actual medical diagnosis. Always consult with qualified healthcare professionals for medical decisions.

• The model is trained on specific types of chest X-ray images
• Performance may vary with different image qualities and formats
• Results should be validated by medical professionals
• The system is not a replacement for professional medical diagnosis

This project is licensed under the MIT License - see the LICENSE file for details.

• YOLOv8: Ultralytics for the excellent object detection framework
• Streamlit: For the powerful web application framework
• Medical Dataset: Contributors to the chest X-ray dataset
• Open Source Community: For various supporting libraries and tools

Built with ❤️ using Streamlit and YOLOv8