DisasterCNN

A simple Streamlit app that predicts disaster types from images using a convolutional neural network.

Overview

This project loads a pre-trained TensorFlow model (model.h5) and runs a web interface using Streamlit. The app classifies uploaded images into one of four disaster categories:

• Cyclone
• Earthquake
• Flood
• Wildfire

The model was trained using a Kaggle dataset for disaster image classification.

Files

• main.py - Streamlit application that loads the model and performs prediction on uploaded images.
• model.h5 - Primary trained TensorFlow model used by the app.
• disastercnn.ipynb - Jupyter notebook for experimentation, exploration, or model training.

Requirements

• Python 3.8+
• TensorFlow
• Streamlit
• Pillow
• NumPy

Install

pip install tensorflow streamlit pillow numpy

Run the app

From the project directory:

streamlit run main.py

Then open the local URL shown in the terminal.

Usage

1. Open the Streamlit app.
2. Upload an image of a disaster scene.
3. View the predicted disaster type and confidence score.

Notes

• The app expects input images to be RGB and resizes them to 224x224 pixels.
• The model predictions are normalized and returned with the highest-confidence class.

License

Include a license or attribution if desired.