Skip to content

sanjay035/Sketch2Color-anime-translation

master
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
Code

Files

Permalink
Failed to load latest commit information.

Sketch2Color-anime-translation:

Given a simple line-art anime sketch the model should output a decent colored anime image using Conditional-Generative Adversarial Networks (C-GANs).

Implemented the anime sketch colorization using the reference(s) paper-1 below.

References:

[1] https://arxiv.org/abs/1705.01908

[2] https://machinelearningmastery.com

[3] https://github.com/soumith/ganhacks

Prerequisites:

  • Miniconda
  • Tensorflow 1.15
  • NVIDIA GPU (8G memory) + CUDA cuDNN

Installation:

  • Clone this repo:
> git clone https://github.com/sanjay235/Sketch2Color-anime-translation.git
> cd Sketch2Color-anime-translation
> conda env create -f environment.yml
  • Then activate the environment created and launch the jupyter:
> conda activate Sketch2ColorAnimeTranslation
> jupyter notebook

Dataset:

  • Download and extract the dataset to current directory.
  • Open an another anaconda prompt with same environment activated and run the pre-processing script as below,
> python DataPreprocessing.py

Training and testing:

  • For the training process refer the Sketch2Color_Anime.ipynb notebook.
  • For the evaluation on test data refer the final.ipynb notebook.

C-GAN Architecture:

C-GAN

Tensorboard logs:

Tensorboard

In progress training results:

At epoch 30,

result_epoch_30

At epoch 40,

result_epoch_40

At epoch 43,

result_epoch_43

Prediction on test sketches:

Sample_1

Sample_2

Contribute:

Correction & Contribution are always welcome!! 😃