Pneumonia Detection from Chest X-Ray — A Complete Project Overview

An end-to-end Deep Learning project that utilizes a Custom Convolutional Neural Network (CNN) to detect and classify instances of Pneumonia from chest X-ray images. The project includes a model training script, a confusion matrix visualization script, and a responsive web application built with Streamlit.

🌟 Features

• Deep Learning Model: A custom-built, lightweight CNN model using TensorFlow & Keras.
• Accurate Detection: Classifies chest X-rays into Normal or Pneumonia.
• Interactive Web App: Easy-to-use Streamlit interface for seamless image uploads and real-time predictions.
• Evaluation Metrics: Scripts provided to evaluate model accuracy and generate visually appealing Confusion Matrices with Seaborn.

🛠️ Technologies Used

• TensorFlow & Keras: For designing and training the deep learning model.
• Streamlit: For quickly building an interactive, responsive front-end.
• Seaborn & Matplotlib: For plotting model metrics and matrices.
• NumPy & Pillow (PIL): For image processing and array manipulation.

📂 Project Structure

The project is well organized into the following structure:

• app.py - The Streamlit web application frontend
• train_model.py - Script used to build and train the custom CNN model
• generate_confusion_matrix.py - Script to plot a confusion matrix from test data
• requirements.txt - Python dependencies
• chest_xray/ - Dataset directory
• pneumonia_model_custom.keras - The saved trained model

🚀 Getting Started & How It Works

1. Dataset

Using the Chest X-Ray Images (Pneumonia) dataset, I structured the data securely into training and testing folders with NORMAL and PNEUMONIA categories.

2. Training the Model

I trained a custom Convolutional Neural Network (CNN) from scratch using train_model.py. This evaluates the dataset and outputs a customized .keras model file.

3. Evaluating the Model

Evaluated the final output by generating a classification report and a beautiful Seaborn confusion matrix plot (confusion_matrix_improved.png).

4. Web Application

I built an easy-to-use visual interface via Streamlit. You can upload any X-ray image, and the model provides an instant prediction along with confidence levels.

🧪 Try It Live & View Code

Click here to try the Pneumonia Detection App

View the source code on GitHub

📌 Final Thoughts

This robust end-to-end Deep Learning application allows us to use advanced CNN architectures seamlessly through a web interface. It highlights the vast utility of AI inside medical image analysis architectures.

🔗 Connect with Me
🌐 www.tauqueeralam.com
📱 LinkedIn | GitHub

View a live demo below: