Skip to content

Latest commit

 

History

History
7 lines (5 loc) · 651 Bytes

README.md

File metadata and controls

7 lines (5 loc) · 651 Bytes

This Repository contains a code to a model to train it on a dataset like AnimeFacesDataset. I've implemented DCGAN from the original paper to generate images. I've implemented 4 similar 2-dimensional convolutions for downsampling image from 64 to 1 as discriminator and 4 similar 2-dimensional convolutions for upsampling in generator. Generator and Discriminator have approximately 10M params. All Hyperparameters were taken from paper.

Generator architecture

image

Model Evolution

git with evolution