Color recommendation for vector graphic documents

Official implementation of Color Recommendation for Vector Graphic Documents based on Multi-Palette Representation

Prerequisites

Python:3.8
Poetry: 1.2.*

Setup

Install requirements and run jupyter.

poetry install
poetry run jupyter lab

Quick demo

notebooks/recomm_colors.ipynb: recommend colors for multiple palettes in a design

Trained model of color prediction are in trained_models/.
Json files for test are pre-created in data/model_test_input/crello_samples/.

You can train a color model on a notebook notebooks/train_model.ipynb. We recommended GPU resources to train this model (e.g. Tesla T4 * 1). You can also create a json file for test from crello dataset on a notebook notebooks/create_json_file.ipynb.

Data

data/training_data/metadata_colors: extracted color palettes for Image-SVG-Text elements from Crello-dataset-v1 (the lastest Crello-dataset)

data/training_data/data_bert/data_color: color corpus of train, validation, and test dataset, and color vocabulary from train dataset

data/trained_models: trained model for color recommendation

model_test_input: json sample files for testing the results of color recommendation

Citation

@misc{Qiu_2022,
  author = {Qiu Qianru, Wang Xueting, Otani Mayu, and Iwazaki Yuki},
  title = {Color Recommendation for Vector Graphic Documents based on Multi-Palette Representation},  
  doi = {10.48550/ARXIV.2209.10820},  
  url = {https://arxiv.org/abs/2209.10820},  
  year = {2022},
}