PyTorch implementation of NeurIPS 2021 paper "Spatial-Temporal Super-Resolution of Satellite Imagery via Conditional Pixel Synthesis" by Yutong He, Dingjie Wang, Nicholas Lai, William Zhang, Chenlin Meng, Marshall Burke, David B. Lobell and Stefano Ermon
Note: Still under construction:)!
pip install -r requirements.txt
To train EAD on Texas housing dataset please run:
python -m torch.distributed.launch --nproc_per_node=8 --master_port=1234 train_3dis.py --path TRAIN_PATH --test_path TEST_PATH --output_dir OUTPUT_DIR
To train EA64 on Texas housing dataset please run:
python -m torch.distributed.launch --nproc_per_node=8 --master_port=1234 train_3dis_attpatch.py --path TRAIN_PATH --test_path TEST_PATH --output_dir OUTPUT_DIR
To train EAD on FMoW-Sentinel2 crop field dataset please run:
python -m torch.distributed.launch --nproc_per_node=8 --master_port=1234 fmow_train_3dis.py --path TRAIN_PATH --test_path TEST_PATH --output_dir OUTPUT_DIR
To train EA64 on FMoW-Sentinel2 crop field dataset please run:
python -m torch.distributed.launch --nproc_per_node=8 --master_port=1234 fmow_train_3dis_attpatch.py --path TRAIN_PATH --test_path TEST_PATH --output_dir OUTPUT_DIR
（FMoW-Sentinel Crop Field Dataset is Still under construction）
If you find our work or datasets useful, please cite the following paper:
@inproceedings{he2021spatial,
title={Spatial-Temporal Super-Resolution of Satellite Imagery via Conditional Pixel Synthesis},
author={He, Yutong and Wang, Dingjie and Lai, Nicholas and Zhang, William and Meng, Chenlin and Burke, Marshall and Lobell, David B. and Ermon, Stefano},
year={2021},
month={December},
abbr={NeurIPS 2021},
booktitle={Neural Information Processing Systems},
}
If you use the FMoW-Sentinel Crop Field dataset, please also cite the original Functional Map of the World paper in addition to ours.
The code is based on the styleganv2 pytorch implementation and CIPS pytorch implementation.
Nvidia-licensed CUDA kernels (fused_bias_act_kernel.cu, upfirdn2d_kernel.cu) are for non-commercial use only.