https://github.com/jleinonen/geogan

Style-based GAN for geophysical fields
https://github.com/jleinonen/geogan

atmospheric-science climate-science deep-learning generative-adversarial-network machine-learning

Last synced: 3 months ago
JSON representation

Style-based GAN for geophysical fields

• Host: GitHub
• URL: https://github.com/jleinonen/geogan
• Owner: jleinonen
• License: mit
• Created: 2019-07-01T12:21:27.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2020-05-07T23:42:36.000Z (almost 5 years ago)
• Last Synced: 2024-01-29T16:27:48.799Z (about 1 year ago)
• Topics: atmospheric-science, climate-science, deep-learning, generative-adversarial-network, machine-learning
• Language: Python
• Size: 24.4 KB
• Stars: 11
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Style-based GAN for geophysical fields

This is a reference implementation of a style-based generative adversarial network (GAN) designed to generate geophysical fields. This code supports a paper published in the [Proceedings of the 9th International Workshop on Climate Informatics: CI 2019](http://dx.doi.org/10.5065/y82j-f154). The paper was selected for the [Best Paper Award](https://sites.google.com/view/climateinformatics2019/accepted-submissions) at the meeting.

You might also be interested in [this paper](https://doi.org/10.1029/2019GL082532) ([code](../../../cloudsat-gan/)) or in [my post on GANs in the atmospheric sciences](https://jleinonen.github.io/2019/06/06/gans-atmos.html).

## Obtaining the data

Download the data from https://doi.org/10.7910/DVN/ZDWWMG and follow the instructions there.

## Obtaining the trained network

Get the trained weights from [this release](../../releases/download/v0.1-data/mch_gan.zip). Unzip the file, preferably into the `models` directory.

## Running the code

For training, you'll want a machine with a GPU and around 32 GB of memory (the training procedure for the radar dataset loads the entire dataset into memory). Running the pre-trained model should work just fine on a CPU.

You may want to work with the code interactively; in this case, just start a Python shell in the `geogan` directory.

If you want the simplest way to run the code, the following two options are available. You may also want to look at what `main.py` does in order to get an idea of how the training flow works.

### Producing plots

You can replicate the plots in the paper by going to the `geogan` directory and using

```

python main.py plot --data_file= --weights_root=

```

where `` is the path to the training data and `` is the path and prefix of the stored weights. For example, if you unzipped the pre-trained weights to the `models` directory, you should use `--weights_root=../models/mch_gan`.

### Training the model

Run the following to start the training:

```

python main.py train --data_file= --weights_root=

```

where the parameters are as above. This will run the training loop until terminated (with e.g. ctrl-C) and save the weights after each 100 batches (overwriting the original weights, so save the weights under a different name if you want to avoid this).