https://github.com/pd-mera/bw-anime-arts-colorization

Using gray image and user's input colors colorize black & white Anime Arts
https://github.com/pd-mera/bw-anime-arts-colorization

Last synced: 3 months ago
JSON representation

Using gray image and user's input colors colorize black & white Anime Arts

• Host: GitHub
• URL: https://github.com/pd-mera/bw-anime-arts-colorization
• Owner: PD-Mera
• Created: 2022-12-08T01:54:14.000Z (over 2 years ago)
• Default Branch: master
• Last Pushed: 2023-02-02T14:35:28.000Z (over 2 years ago)
• Last Synced: 2025-01-08T19:46:57.012Z (5 months ago)
• Language: Python
• Homepage:
• Size: 3.33 MB
• Stars: 3
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Colorize Black & White Anime Arts with User's input colors

Colorize Black & White Anime Arts with user's input colors

*I try to add Global Hint Networks but results is not good as I imagine*

**No Hint**

| Gray | Fake | Real |

| --- | --- | --- |

| ![gray](./assets/imgs/gray.jpg "gray") | ![fake_nohint](./assets/imgs/no_hint/fake.jpg "fake_nohint") | ![real_nohint](./assets/imgs/test_img.jpg "real") |

**Local Hint**

| Gray | Local Hint | Fake | Real |

| --- | --- | --- | --- |

| ![gray](./assets/imgs/gray.jpg "gray") | ![local_hint](./assets/imgs/local/hint.jpg "local_hint") | ![fake](./assets/imgs/local/fake.jpg "fake") | ![real](./assets/imgs/test_img.jpg "real") |

**Global Hint**

| Gray | Global Hint | Fake | Real |

| --- | --- | --- | --- |

| ![gray](./assets/imgs/gray.jpg "gray") | ![local_hint](./assets/imgs/global/hint.jpg "local_hint") | ![fake](./assets/imgs/global/fake.jpg "fake") | ![real](./assets/imgs/test_img.jpg "real") |

## Environments and Dependencies

- Python 3.8.16

Install requirements

``` bash

pip install -r requirements.txt

```

## Data

Prepare a `./data/` directory contain all subfolder of images used for training in format

``` folder

data/

    |-- train/

    |   |-- class 1/

    |   |   |-- img 1

    |   |   |-- img 2

    |   |   `-- ...

    |   |-- class 2/

    |   |   |-- img 1

    |   |   |-- img 2

    |   |   `-- ...

    |   `-- ...

    |-- train_small/

    |   `-- small_set/

    |       |-- img 1

    |       |-- img 2

    |       `-- ...

    `-- val/

        `-- classname/

            |-- img 1

            |-- img 2

            `-- ...

```

## Easy Use

For easy use, simply download pretrained weight from [here](https://drive.google.com/file/d/1HrMR3Holk2TSKQUCbmYHCZL5rDqu2Iag/view?usp=share_link) (10 * loss for global) or [here](https://drive.google.com/file/d/1XjLuJLB4ydWMvQq5-hhocO6DIy8vJH8F/view?usp=sharing) (100 * loss for global), modify config in `config.py` and run 

``` bash

python infer.py

```

## Training

You can update hyperparameter in `train.sh` and training a model from scratch using

``` bash

bash train.sh

```

## Valid

Update hyperparameter in `test.sh` and valid a model using

``` bash

bash test.sh

```

## TODO

- [x] Add Global Hints

- [ ] Improve Global Hints

## Acknowledgments

This coding is heavily followed by [colorization-pytorch](https://github.com/richzhang/colorization-pytorch)'s repo.