Deep-PHURIE: Hurricane Intensity Estimation with Deep Learning

Overview

Deep-PHURIE is a deep learning-based model for hurricane intensity estimation from infrared satellite imagery. It utilizes convolutional neural networks (CNNs) to predict hurricane intensity based on provided features.

Dataset

The model is trained on the TCIR-ALL dataset from 2017. The dataset contains infrared and passive microwave (PMW) data, with corresponding hurricane intensity labels (Vmax).

Anaconda Environment

Ensure you have Anaconda installed. Create and activate the conda environment using the provided environment.yml file:

conda env create -f anaconda_env.yml
conda activate cyclone_env

Exploratory Data Analysis (EDA)

Explore the dataset to understand its structure and characteristics:

# Run EDA
python EDA&Preprocessing.ipynb

Model Architecture

AlexNet

The model architecture is based on AlexNet and consists of convolutional layers, max-pooling layers, batch normalization, and fully connected layers.

DeepPHURIE Model

Additionally, a modified DeepPHURIE model is implemented for hurricane intensity prediction using a different architecture.

Training

The models are trained using k-fold cross-validation, and training results are visualized using matplotlib.

# Train the models
python Model_Training.ipynb

License

This project is licensed under the MIT License - see the LICENSE file for details.