ID-GAN PyTorch Implementation

Reference paper: https://arxiv.org/abs/2001.04296

Auther's implementation: https://github.com/1Konny/idgan.git

---

Training

• Data Set: CelebA data set is used for model training
• Stage 1: VAE is trained for 1e6 iteration
• Stage 2: Resnet Generator & Discriminator are trained for 3e5 iteration
• trained on 1xA100 (40 GB SXM4) for 28hrs

---

Project Files and Folders

• /d_chkpt - Folder where 'Discriminator' model check points to be saved
  • getModelPkl.py: used to download final check point for the trained model. Command: python getModelPkl.py
• /g_chkpt - Folder where 'Generator' model check points to be saved
  • getModelPkl.py: used to download final check point for the trained model. Command: python getModelPkl.py
• /vae_chkpt - Folder where 'VAE' model check points to be saved
  • getModelPkl.py: used to download final check point for the trained model. Command: python getModelPkl.py
• /outputs - Folder where the generated images are saved
• /ID-GAN.ipynb - Jupyter notebook for the paper implementation.
• /requirements.txt - Needed python libraries