Chest X-Ray Classification using EfficientNetB0

This repository contains a TensorFlow/Keras implementation of a deep learning model to classify chest X-ray images into Normal and Pneumonia categories using the EfficientNetB0 architecture. The model leverages transfer learning and data augmentation to improve accuracy and generalization.

---

Table of Contents

• Overview
• Dataset
• Model Architecture
• Installation
• Training
• Evaluation
• Results
• Saving & Loading Model

---

Overview

This project aims to build a high-performance chest X-ray classifier that can differentiate between healthy lungs and pneumonia-affected lungs. The model uses EfficientNetB0 pre-trained on ImageNet and fine-tunes a few dense layers for binary classification.

Features:

• Transfer learning with EfficientNetB0
• Data augmentation for better generalization
• Early stopping and learning rate reduction on plateau
• Confusion matrix and classification report for evaluation

---

Dataset

Kaggle link: https://www.kaggle.com/datasets/yusufmurtaza01/chest-xray-pneumonia-balanced-dataset The dataset should have the following structure:

data/
├── train/
│   ├── Normal/
│   └── Pneumonia/
└── test/
    ├── Normal/
    └── Pneumonia/

• train: training and validation split (80-20)
• test: independent test set
• Images resized to 224x224 pixels

Note: Adjust the paths in the code according to your dataset location.

---

Model Architecture

The model uses EfficientNetB0 as a backbone:

• Pre-trained EfficientNetB0 (weights=imagenet, include_top=False)

• Global Max Pooling

• Dense layers:

  • 512 units + ReLU + Dropout 0.5
  • 256 units + ReLU + Dropout 0.4
  • Output: 2 units (softmax for Normal/Pneumonia)

Training Strategy:

• Optimizer: Adam (learning_rate=0.001)
• Loss: Categorical Crossentropy
• Metrics: Accuracy
• Callbacks: EarlyStopping and ReduceLROnPlateau

---

Installation

Clone the repository and install required packages:

git clone https://github.com/Escobar-12/Chest_X-Ray_Pneumonia_Classification.git
cd chest-xray-classification
pip install -r requirements.txt

---

Requirements

• Python 3.10
• TensorFlow 2.10.1
• NumPy 1.26.4
• Pandas
• Matplotlib
• Seaborn
• Scikit-learn
• CUDA 11.2 and cuDNN 8.1 (for GPU support)
• Conda (recommended for environment management)

---

Installation if using GPU for training

1. Create a conda environment:

conda create -n tf_gpu python=3.10
conda activate tf_gpu

2. Install GPU dependencies (if using GPU):

conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0

3. Install Python libraries:

pip install tensorflow==2.10.1 numpy==1.26.4 pandas matplotlib seaborn scikit-learn

3. Visualize training results and evaluate the model.

---

Training

• Training is performed with a batch size of 32
• Image preprocessing uses EfficientNetB0 preprocessing (preprocess_input)
• Data augmentation includes random brightness and contrast

Example plot outputs:

• Training vs Validation Accuracy
• Training vs Validation Loss

---

Evaluation

After training, the model evaluates on the test set:

• Confusion Matrix
• Classification Report (Precision, Recall, F1-Score)

Example:

y_pred_probs = model.predict(test)
y_pred = np.argmax(y_pred_probs, axis=1)
y_true = ...

cm = confusion_matrix(y_true, y_pred)
sns.heatmap(cm, annot=True, fmt="d", xticklabels=classes, yticklabels=classes)

---

Results

• Test Accuracy: ~97% (may vary depending on dataset)
• Confusion matrix and classification report included in scripts

---

Saving & Loading Model

Save model weights:

model.save_weights('./models/chest_xray_model_weights.h5')

Load weights:

model.load_weights('./models/chest_xray_model_weights.h5')

---

License

This project is licensed under the MIT License.