Skip to content

Normalize rasterization line-drawings to uniform width using deep learning.

License

Notifications You must be signed in to change notification settings

hepesu/LineNormalizer

Repository files navigation

LineNormalizer

Overview

Normalize rasterization line-drawings to uniform width using deep learning with model from Smart Inker.

This model can serve as line-drawings preprocessor for LineRelifer. Line-drawings can be normlized to an intermediate representation and then be used as training data or input for it. Also by using this method, we can achieve uniform line width during scaling up or down the rasterization line-drawings, which is a feature of vector line-drawings. The train data is generated by code, so you can get model for any width easily.

Dependencies

  • Keras2 (Tensorflow1 backend)
  • Pytorch
  • OpenCV3
  • CairoSVG

Usage

  1. Set up directories.

  2. Download the model from release and put it in the same folder with code.

  3. Run predict.py for prediction. Run model{NUM}.py for train.

Files with name starts with pytorch are Pytorch version.

Models

Models are licensed under a CC-BY-NC-SA 4.0 international license.

Keras

  • model_180913

Pytorch

  • model_200801

From Project HAT by Hepesu With ❤️