Autoencoder-Based-Abnormality-Detection-in-Bone-X-Ray-images-Using-MURA-Dataset

An autoencoder-based model to detect abnormalities in bone X-rays using the MURA dataset. We preprocess images (96×96), extract features using ResNet50, and address class imbalance with SMOTETomek. Our model achieves 94% accuracy, aiding radiologists in diagnosis.

📂 Project Structure

• MURA.ipynb – Contains the proposed model and experiment details.
• comparison.ipynb – Includes results from different standard classifiers using the same dataset.
• test.ipynb – Code to test the trained model on any test.png image.
• encoder.h5 – Pre-trained ResNet50 encoder for feature extraction.
• classifier.h5 – The proposed autoencoder-based classifier model.