This repository contains the code and resources for the STORM project, a deep learning model for precipitation nowcasting based on weather radar data. The model was developed as part of the Deep Learning Workshop at Tel Aviv University.

The project aims to predict future radar-based precipitation frames using a sequence of past frames. The core of the project is the STORM model, a U-Net style autoencoder with a ConvLSTM bottleneck, designed to capture both spatial and temporal dynamics.

STORM-Nowcasting/
├── data_creation/      # Scripts to download and process the raw radar data
├── models/             # Contains the STORM model architecture definition
├── training/           # Scripts and configs for training the model and running sweeps
├── evaluation/         # Scripts for quantitative and qualitative evaluation
├── notebooks/          # Jupyter notebook for demonstration and analysis
└── saved_weights/       # Directory to store trained model weights