https://github.com/bennourines/pneumonia-detection-cnn

CNN model for pneumonia detection from chest X-ray images, fine-tuned with ResNet50 and achieving 94% accuracy on ChestX-ray dataset.
https://github.com/bennourines/pneumonia-detection-cnn

cnn computer-vision deep-learning image-classification machine-learning resnet-50 tensorflow

Last synced: 20 days ago
JSON representation

CNN model for pneumonia detection from chest X-ray images, fine-tuned with ResNet50 and achieving 94% accuracy on ChestX-ray dataset.

• Host: GitHub
• URL: https://github.com/bennourines/pneumonia-detection-cnn
• Owner: bennourines
• Created: 2025-11-12T01:37:33.000Z (8 months ago)
• Default Branch: main
• Last Pushed: 2025-11-12T01:41:32.000Z (8 months ago)
• Last Synced: 2025-11-12T03:19:58.248Z (8 months ago)
• Topics: cnn, computer-vision, deep-learning, image-classification, machine-learning, resnet-50, tensorflow
• Language: Jupyter Notebook
• Homepage:
• Size: 3.32 MB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# 🩺 Pneumonia Detection from Chest X-Ray Images

### 🎓 Project — ESPRIT (2024)

**Author:** Inès Bennour  

**Institution:** ESPRIT, Tunisia  

**Year:** 2024  

---

## 🧠 Project Overview

This project focuses on building a **Convolutional Neural Network (CNN)** for the **automatic detection of pneumonia** from chest X-ray images.  

The goal is to assist radiologists in early diagnosis by leveraging **deep learning** and **computer vision**.

---

## 🚀 Objectives

- Design a CNN architecture capable of classifying X-ray images into **Pneumonia** vs **Normal**.

- Enhance model performance through:

  - **Data Augmentation**

  - **Fine-tuning of the ResNet50** pre-trained model.

- Achieve high diagnostic accuracy on real-world medical data.

---

## 🧩 Dataset

- **Dataset Used:** [ChestX-ray Pneumonia Dataset (Kaggle)](https://www.kaggle.com/datasets/paultimothymooney/chest-xray-pneumonia)

- **Classes:**  

  - `PNEUMONIA`  

  - `NORMAL`

- **Dataset Split:**  

  - 70% training  

  - 20% validation  

  - 10% testing  

---

## 🏗️ Model Architecture

### 1️⃣ Baseline CNN

- Custom CNN with convolutional + pooling layers.

- Activation: ReLU  

- Optimizer: Adam  

- Loss: Binary Crossentropy  

### 2️⃣ Fine-Tuned ResNet50

- Pre-trained on **ImageNet**.

- Top layers replaced with custom dense layers.

- Data augmentation applied (rotation, zoom, horizontal flip).

- Fine-tuning performed on last layers for optimal performance.

---

## 📊 Results

| Model                | Accuracy | Precision | Recall | F1-Score |

|----------------------|-----------|------------|--------|-----------|

| Baseline CNN         | 87%       | 85%        | 88%    | 86%       |

| **ResNet50 (Fine-Tuned)** | **94%** | **93%** | **95%** | **94%** |

✅ **Final Model Accuracy: 94%**

---

## 🛠️ Technologies & Tools

- **Python**

- **TensorFlow / Keras**

- **OpenCV**

- **Matplotlib / Seaborn**

- **NumPy / Pandas**

- **Jupyter Notebook**

---