PyTorch Project to Build a GAN Model on MNIST Dataset

This project focuses on training and comparing Vanilla GAN and WGAN using the MNIST dataset to generate realistic images. The insights gained contribute to understanding GAN architecture and its gaming, medical imaging, and digital content creation applications.

• Successfully trained and compared Vanilla GAN and WGAN using the MNIST dataset.

• Generated realistic images with WGAN outperforming Vanilla GAN in quality.

• Evaluated image quality using FID and Inception Score, highlighting WGAN's superiority.

• Visualized real vs. fake images for easy comparison.

• Calculated FID and Inception Score to assess image similarity and diversity.

• Demonstrated WGAN's stability in training and image generation.

• Gained insights into the impact of GAN architecture on image quality.

• Contributed to understanding GAN performance with key evaluation metrics.

• GANs can generate realistic assets for gaming, medical imaging, and e-commerce.

• Improved image generation benefits digital marketing and content creation industries.