Building Upon White-Box Cartoonization as Described in CVPR2020 Paper

Original Paper and Source

Learning to Cartoonize Using White-box Cartoon Representations
Output examples can be found at both of the following links: original project page | paper

Usage

Installation

Prerequisites (Windows):

• NVIDIA CUDA and CuDNN
• MSVC 2015 (found within Microsoft C++ Build Tools)
• Python 3.6 is the latest version compatible with the required tensorflow versions

> python36 -m venv .\wbcvenv
> .\wbcvenv\Scripts\activate
> python36 -m pip install pip
> python36 -m pip install tensorflow==1.12.0
> python36 -m pip install tensorflow-gpu==1.12.0
> python36 -m pip install scikit-image==0.14.5
> python36 -m pip install opencv-python
> python36 -m pip install tqdm

Inference with Pre-trained Model

• Store test images in /test_code/test_images
• Run ./cartoonize.py
• Results will be saved in /test_code/cartoonized_images

Train

• Place your training data in corresponding folders in /dataset
• Run pretrain.py, results will be saved in /pretrain folder
• Run train.py, results will be saved in /train_cartoon folder
• Codes are cleaned from production environment and untested
• There may be minor problems but should be easy to resolve
• Pretrained VGG_19 model can be found at following here (link provided by SystemErrorWang).

Datasets

• Due to copyright issues, we cannot provide cartoon images used for training
• However, these training datasets are easy to prepare
• Scenery images are collected from Shinkai Makoto, Miyazaki Hayao and Hosoda Mamoru films
• Clip films into frames and random crop and resize to 256x256
• Portrait images are from Kyoto animations and PA Works
• We use this repo to detect facial areas
• Manual data cleaning will greatly increace both datasets quality