Using FFHQ alignement method with mediapipe (or dlib) as landmarks detection method.
Launch the following script:
python generate_image.py
- --InFolder: Input images directory
- --OutFolder: Output directory for generated images
- --method: landmark library. So far, supported are 'mediapipe' and 'dlib'
- --no_padding: Do not us padding.
- --rotate_level. Do not rotate face.
-
- --size_output. Image output size. Default is 1024
Script is able to manage multi faces in a single image.
Excellent stuff from this paper
@articleFreeStyleGAN2021, author = Thomas Leimkühler and George Drettakis, title = FreeStyleGAN: Free-view Editable Portrait Rendering with the Camera Manifold, booktitle = ACM Transactions on Graphics (SIGGRAPH Asia), publisher = ACM, volume = 40, number = 6, year = 2021, doi = 10.1145/3478513.3480538
-
Photo by Batel Studio on Unsplash
-
Photo by Mubariz Mehdizadeh on Unsplash
-
Photo by Michael Afonso on Unsplash
-
Photo by Thought Catalog on Unsplash