This repository presents the implementation of a Spectral Normalization Generative Adversarial Network (SN GAN) using PyTorch for image generation. The notebook SN_GAN_MNIST_PyTorch.ipynb details the step-by-step implementation and training process.
-
Epoch GIFs: This directory contains GIFs illustrating the generated images at each epoch during the 20-epoch training process. These visuals offer insights into the model's learning progression.
-
Generator Images: Explore this folder to find the images generated by the model for each epoch. These images showcase the evolution of the generator's output over the course of training.
-
SN_GAN_MNIST_PyTorch.ipynb: The main notebook where the SN GAN is implemented. It provides a comprehensive guide through the code, training dynamics, and evaluation metrics.
The training journey encompasses the model's understanding of contextual information, addressing challenges such as mode collapse, and achieving stability. The SSIM metric is utilized for evaluation, resulting in an impressive average score of 0.94, affirming the model's proficiency in generating realistic images.
Feel free to explore the GIFs and generated images to witness the model's evolution. The notebook serves as a valuable resource for understanding the implementation details and training nuances of the SN GAN for MNIST dataset.
Happy exploring and generating!
