Frequency Estimation Using Complex-Valued Shifted Window Transformer

The code accompanying our paper currently in preprint at arXiv for complex-valued shifted window (Swin) transformer-based spectral super-resolution.

Publication and Citation

If you appreciate our work, please cite our work as

@article{smith2023frequency,
	title        = {Frequency Estimation Using Complex-Valued Shifted Window Transformer},
	author       = {Smith, J. W. and Torlak, M.},
	year         = 2023,
	month        = sep,
	journal      = {arXiv preprint arXiv:2305.02017}
}

Getting Started

This is an example of how you may give instructions on setting up your project locally. To get a local copy up and running follow these simple example steps.

Prerequisites

This is an example of how to list things you need to use the software and how to install them.

PyTorch

PyTorch must be install using the CPU/GPU configuration desired prior to attempting to run the code.
requirements.txt
```
pip install -r requirements.txt
```

Training

Both SwinFreq and CVSwinFreq can be easily trained from scratch using the included tools.

Training SwinFreq
```
python train.py --model swinfreq
```
Training CVSwinFreq
```
python train.py --model cvswinfreq
```

Reproducing Paper Experiments

The experiments shown in the paper can be reproduced by calling paper_experiments. The trained models are contained in the folder saved/models/. sh python paper_experiments.py After running the experiments paper_results/make_figures.m can be run in MATLAB to reproduce the figures used to create the figures and save them to both .fig and .png files.

Note: experiment 4, showing a rotating target and demonstrating the improved resolution capacity of the proposed methods, was not used in the paper due to space constraits.

License

Distributed under the GPL-3.0 License. See LICENSE.txt for more information.

Contact

Josiah W. Smith - josiah.radar@gmail.com

Project Link: https://github.com/josiahwsmith10/spectral-super-resolution-swin