Object Detection & Classification Model - MURA Dataset

📌 Overview

This repository implements a deep learning solution for detecting abnormalities in musculoskeletal radiographs using the MURA (MUsculoskeletal RAdiographs) dataset.

Beyond standard binary classification (Normal vs. Abnormal), this project focuses on Explainable AI (XAI). It includes an explainer module that generates heatmaps to visualize the specific regions of the X-ray that led to the model's decision, effectively providing weakly-supervised object detection and localization of fractures or anomalies.

📂 Dataset

The project utilizes the MURA Dataset, one of the largest public radiographic image datasets, released by the Stanford ML Group.

• Task: Binary Classification (Normal vs. Abnormal)
• Body Parts Covered: Elbow, Finger, Forearm, Hand, Humerus, Shoulder, Wrist.
• Dataset Link: link

🚀 Key Features

• Robust Classification: A CNN-based architecture designed to handle the variability in X-ray imaging.
• Model Explainability: Uses techniques (likely Grad-CAM or Saliency Maps) to generate heatmaps, highlighting the "area of interest" where the abnormality is detected.
• Performance Tracking: Visualizes training progress including accuracy and loss curves.

📁 Repository Structure

File / Folder	Description
XRayClassification.py	Main script containing the model architecture, training loop, and evaluation logic.
explainer.py	Script responsible for generating visual explanations (heatmaps) to interpret model predictions.
heatmaps/	Directory where generated heatmap images are saved.
training_history.png	Plot showing the training and validation loss/accuracy over epochs.
sample_predictions.png	Example output image showing input X-rays and model predictions.

🛠️ Installation & Requirements

1. Clone the repository:

   git clone [https://github.com/Wolfram-St/Object-Detection-Classification-Model-MURA-Dataset-.git](https://github.com/Wolfram-St/Object-Detection-Classification-Model-MURA-Dataset-.git)
   cd Object-Detection-Classification-Model-MURA-Dataset-

2. Install dependencies: (Note: Ensure you have Python installed. The specific requirements depend on the libraries used in the scripts, but generally include the following)

   pip install numpy pandas matplotlib opencv-python scikit-learn
   # specific deep learning framework (install the one used in XRayClassification.py)
   # pip install torch torchvision  # If using PyTorch
   # pip install tensorflow keras   # If using TensorFlow/Keras

💻 Usage

1. Train the Model

To start training the classification model on the MURA dataset:

python XRayClassification.py