1002753959/auto_painter

Recently, realistic image generation using deep neural networks has become a hot topic in machine learning and computer vision. Such an image can be generated at pixel level by learning from a large collection of images. Learning to generate colorful cartoon images from black-and-white sketches is not only an interesting research problem, but also a useful application in digital entertainment. In this paper, we investigate the sketch-to-image synthesis problem by using conditional generative adversarial networks (cGAN). We propose a model called auto-painter which can automatically generate compatible colors given a sketch. Wasserstein distance is used in training cGAN to overcome model collapse and enable the model converged much better. The new model is not only capable of painting hand-draw sketch with compatible colors, but also allowing users to indicate preferred colors. Experimental results on different sketch datasets show that the auto-painter performs better than other existing image-to-image methods.

Auto_painter

News

We have released our dataset for public using. The dataset can be downloaded through following links:

Sketch-image pairs: https://cloudstor.aarnet.edu.au/plus/s/rMSBYCjEZJ70ab2

Sketch with control color blocks: https://cloudstor.aarnet.edu.au/plus/s/ixj8XS0rMmUqq0Z

Orginal README

It is the original implementation of the journal article: Auto-painter: Cartoon image generation from sketch by using conditional Wasserstein generative adversarial networks https://www.sciencedirect.com/science/article/pii/S0925231218306209?via%3Dihub

This project mean to make an end-to-end network for the sketch of cartoon to have color automatically.

More results can be seen here: https://irfanicmll.github.io/work-page/

Try our demo here: http://103.202.133.77:10086/

Since the lab's server has temporarily expired, the demo is now unavailable. You can see the demo video and train your own model.

New model has been updated!~ The performance is much better than in the orginal paper! See the demo video:https://youtu.be/g9rf-YFGgbg

Have a try~

The pre-train model can be download here: https://cloudstor.aarnet.edu.au/plus/s/LvyREKsiaH47Aa6

My homepage: https://irfanicmll.github.io/

Welcome to contact me~

Dependencies

python3.5

tensorflow1.4

Vgg model from:https://github.com/machrisaa/tensorflow-vgg(optional, if you use the loss_f)

Data

Color images: Collected on the Internet

Sketch: Generated from the preprocessing/gen_sketch/sketch.py

Quick start

Put you orginal data in the folder preprocessing/gen_sketch/pic_org

Run the sketch.py and you will get the training set in the preprocessing/gen_sketch/pic_sketch folder

Download the pre-train weight of Vgg16, and put the model and the pretrian weight uder the folder of training&test/my_vgg

Run the training command as:

python auto-painter.py --mode train --input_dir $TRAINING_SET --output_dir $OUTPUT --checkpoint None

Run the testing command as:

python auto-painter.py --mode test --input_dir $TESTING_SET --output_dir $OUTPUT_TEST --checkpoint $OUTPUT