README.md

Turn Your Photos Into A Comic Book

Table of contents

1. Directory Structure
2. Dataset
3. Models
4. Results

Directory Structure

./comicface-ai/*
  ├─ data - # The training dataset will go here. A few sample images are added for reference
  ├─ src - # Model architecture, data loading, and other utility scripts are saved here
  ├─ notebooks - # The notebooks show an end-to-end pipeline to train and infer the models. The notebook names are self-explanatory
  ├─ saved_models - # Trained models

Dataset

The dataset is sourced from Kaggle and contains 10,000 pairs of faces and their comic version. Each image is of 1024x1024 dimensions.

Models

Two different model architectures are tried for training: Convolutional Variational Autoencoder and Pix2Pix GAN.

Convolutional VAE

Model Architecture

Convolutional VAE Training

Pix2Pix GAN

Model Architecture

Pix2Pix GAN Training

Results

Convolutional VAE Sample Results

Pix2Pix GAN Sample Results