sudheerachary / Manga_Colorization

cGAN-based Manga Colorization Using a Single Training Image.

Manga Colorization

• models/ consists the cGAN models & visualizations.

  • run train.py to train on images stored in datasets/train.
  • run test.py to generate colorization for images in datasets/test.
  • datasets/generated_Images contains generated images and trained model weights.
  • datasets/test contains testing images.
  • datasets/train contains/put dataset for training.
  • datasets/validation contains validation set of images for training.

• Segmentation and post-processing

  • segmentation/input_images contains test images
  • segmentation/bw_images contains the black and white images after screentone removal
  • Results contains results for postprocessing and final outputs
  • run segmentation.m in MATLAB to segment the input image (Monochrome Original)

• cGAN + =

a. b. c.

• a - input: monochrome original
• b - output: cGAN output
• c - output: Post Processed output