3d-sr-micrometeorology

This repository contains the source code used in Super-Resolution of Three-Dimensional Temperature and Velocity for Building-Resolving Urban Micrometeorology Using Physics-Guided Convolutional Neural Networks with Image Inpainting Techniques (arXiv, bae).

• Setup
  • Docker Containers
  • Singularity Containers
• Code used in experiments
• Citation

Setup

• The Singularity containers were used for experiments.
• The Docker containers provide the same environments as in the Singularity containers.

Docker Containers

1. Install Docker.
2. Build docker containers: $ docker compose build
3. Start docker containers: $ docker compose up -d

Singularity Containers

1. Install Singularity.
2. Build Singularity containers:
   • $ singularity build -f datascience.sif ./singularity/datascience.sif.def
   • $ singularity build -f pytorch.sif ./singularity/pytorch.def
3. Start singularity containers:
   • The following command is an exmple for local.

export PORT=8888 # your own port
singularity exec --nv --env PYTHONPATH="$(pwd)/pytorch" \
  ./pytorch.sif jupyter lab --no-browser --ip=0.0.0.0 --allow-root --LabApp.token='' --port=$PORT

Code used in experiments

• Data generation for deep learning
  • This code was run using the datascience container on the Earth Simulator.
• CNN training
  • The root directory and config paths are need to be specified.
  • This shell script runs python script.
    • This python script was run using the pytorch container on the Earth Simulator.
• CNN evaluation
  • The CNNs were evaluated using notebooks in the above dir.
  • These notebooks wre run using the pytorch container on a local environment.

Citation

@article{
  title = {Super-resolution of three-dimensional temperature and velocity for building-resolving urban micrometeorology using physics-guided convolutional neural networks with image inpainting techniques},
  journal = {Building and Environment},
  volume = {243},
  pages = {110613},
  year = {2023},
  issn = {0360-1323},
  doi = {https://doi.org/10.1016/j.buildenv.2023.110613},
  url = {https://www.sciencedirect.com/science/article/pii/S0360132323006406},
  author = {Yuki Yasuda and Ryo Onishi and Keigo Matsuda}
}