Evaluate data on baseline using Generate_Profile_Images_baseline. Select appropriate results and evaluate on Face_Verification.
Generating fake profile images that can be verified with your own face but look as different as possible.
All files can be run locally using jupyter.
- Preprocess images from your personal dataset by running Data_Preprocessing.ipynb.
- Set up the face verification system by running Face_Verification.ipynb. Save the generated database.pkl file for later use.
- To test baseline use Generate_Profile_Images_baseline.ipynb. This is based on FreezeD.
- To test on starGAN v2 use starGAN_v2.ipynb.
- Evaluate selected generated images in Face_Verification.ipynb under Experiments by loading the previously downloaded database.pkl file.
- To experiment with a naive Freeze-D implementation for starGAN, replace the solver.py file in the core folder (after downloading the repository) with the solver.py file in this repository.