Anime-Colorization-Machine

User-interactive colorization of anime sketch image

Prerequisites

pytorch

torchvision

numpy

openCV2

PyQT5

opencv_transforms (For training) (You can simply install this by pip install opencv_transforms)

matplotlib (For training)

Results

Demo video

Dataset

1. Taebum Kim, "Anime Sketch Colorization Pair", https://www.kaggle.com/ktaebum/anime-sketch-colorization-pair

2. 68K illustrations crawled from web

Usage

1. Download model weight from here and unzip on src/model/checkpoint

2. python main.py

3. Load an image. The image will be automatically converted to sketch-style image.

4. Draw scribbles and press 'colorize' button.

5. Save the colorized image.

Training details

Parameter	Value
Learning rate	2e-4
Batch size	2
Iteration	150K
Optimizer	Adam
(beta1, beta2)	(0.5, 0.999)
Data Augmentation	RandomResizedCrop(512) RandomHorizontalFlip()
HW	CPU : Intel i5-8400 RAM : 16G GPU : NVIDIA GTX1060 6G
Training Time	About 1.37s per iteration (About 50 hours for 150K iterations)