TensorFlow template for EarthNet2021 dataset

This git repository contains the python code for the TensorFlow template of EarthNet2021.

This work is based on the Tensorflow implementation of Stochastic Adversarial Video Prediction by Alex Lee et al. (ArXiv)

Content of the Repository

This repository contains two entry points for training and testing, namely 'train_earthnet.py' and 'generate_earthnet.py'.

This TensorFlow template uses old TF 1.15. Are you developing TF 2.0 work for EartNet? Please contribute to the EarthNet2021 intercomparison toolkit! :)

Getting started

To get started clone the repository or use it as a github template.

Requirements

We tested the code using Python 3.6, Tensorflow 1.9, CUDA Version 9.0.176 and CUDNN 7.4.2

Make sure the python dependencies in Requirements are installed. Alternatively create a conda environment using conda.yml settings.

General workflow

numpy matplotlib scipy tensorflow-gpu==1.15