DiffFace is a project that implements a diffusion model for generating synthetic face images of dark-skinned individuals, for the purpose of improving the diversity of datasets used in facial recognition systems. The project aims to address the underrepresentation of dark-skinned faces in existing datasets, which can lead to biased and inaccurate results in facial recognition applications.
To install the required dependencies, run the following command:
pip install -r requirements.txtTo get the training data, run the following command:
bash dataset.shTo train the model, use the following command:
python src/train.pyTo sample images from the trained model, use the following command:
python src/sample.pyThe checkpoints for the trained model can be downloaded using the following command:
bash checkpoints/checkpoints.shThe configuration for the model can be found in the config.py file. You can modify the parameters in this file to customize the training and sampling process.
An example of the results generated by the model is shown below:
