Climate Reconstruction via Image Inpainting using Partial Convolutions

[Applied implementation] (https://github.com/naoto0804/pytorch-inpainting-with-partial-conv)

Official implementation is released by the authors.

Note that this is an ongoing re-implementation and is designed for climate reconstructions using numerial model input and output!

This is an unofficial pytorch implementation of a paper, Image Inpainting for Irregular Holes Using Partial Convolutions [Liu+, arXiv2018].

Requirements

• Python 3.6+
• Pytorch 0.4.1+

pip install -r requirements.txt

Usage

Preprocess

• download climate data (e.g. 20CR reanalysis) and preprocess it to hdf5 5x5° (see mask dir). Use anomalies (image normalization is turned off) The dataset should contain data_large, val_large, and test_large as the subdirectories. Don't forget to specify the root of the dataset by --root ROOT when using train.py or test.py

Use the start.sh script to operate the process chain: Train, Fine-Tune, Test

Results

Can be found in Kadow et al. 2020 "Artificial Intelligence Reconstructs Missing Climate Information"