The repository contains Conditional Generative Adversarial Network (CGAN) model, build using Pytorch, to classify and reconstruct the ancient coins.
The dataset been used to train the model includes Hadrian Roman Imperial Coins obtained from publicly available Online Coins of the Roman Empire (OCRE) database.
Dataset-Sample
image size = [128, 128]
number of epochs = 2
generator dimensions = 100
batch size = 128
learning rate = 0.0002
The reconstructed samples could be obversed under this directory. A better resolution synthesis could be achieved using the coin samples scaled greater than 128x128.