PETIT-GAN

This repository contains the source code for the methods and results introced in the paper PETIT: Physically Enhanced Thermal Images Translating GAN, which was published in WACV 2024

Throughout the code/documentation, the term monochromatic and 9000nm are used interchangeably, as the 9000nm channel was the monochromatic thermal modality demonstrated in the paper.

For more details about the research, please checkout our project's website

Setup

• Install the required packages either using either:
  • conda (recommended): conda env create -f environment.yml.
  • pip : pip install -r requirements.txt.
• Run the download.py script to download and unpack all the data and pretrained models. NOTE: The data used here is a fraction of the entire dataset (both in terms of amounts of images and modalities). For the full data set, click here.
• Extract the folders in the downloaded zip to the root directory of the repository.
• After completing the setup, the project's root should have the following structure:

├── data
├── docs
├── models
├── src
├── .gitignore
├── download.py
├── environment.yml
├── inference.ipynb
├── requirements.txt
├── train.ipynb

Train

Run the file train.ipynb in the root directory of the repository.

• The notebook will save the trained models in the path results/train/<CurrentTime>.
• The best model in term of FID will be saved in the path results/train/<CurrentTime>/<NetName>/best.
• The latest model will be saved in the path results/train/<CurrentTime>/<NetName>/latest.

Inference

Run the file inference.ipynb in the root directory of the repository. The notebook will produce a folder named results/transformed/<CurrentTime> containing .png/.npy files of the generated images, depending on the user's specification in the notebook.

Citation

If you use PETIT-GAN's code/paper for your research, please cite using the following BibTex:

@InProceedings{Berman_2024_WACV,
  author    = {Berman, Omri and Oz, Navot and Mendlovic, David and Sochen, Nir and Cohen, Yafit and Klapp, Iftach},
  title     = {PETIT-GAN: Physically Enhanced Thermal Image-Translating Generative Adversarial Network},
  booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
  month     = {January},
  year      = {2024},
  pages     = {1618-1627}
}