GuidingPainter

Official implementation of "Guiding Users to Where to Give Color Hints for Efficient Interactive Sketch Colorization via Unsupervised Region Prioritization" (WACV 2023). [arxiv]

Prerequisites

  • pytorch
  • tensorboard
  • pytorch lightning
  • albumentations

Important options

  • phase: train or test
  • processed_dir: directory which contains datasets
  • train_dir: drectory in which checkpoints and logs are saved
  • test_dir: drectory in which test results are saved
  • dataset: the name of dataset
  • gpus: gpus which are used for training
  • test_gpu: single gpu which is used for testing
  • checkpoint: filename of checkpoint

Training example

python main.py \
--dataset {dataset name} \
--processed_dir {processed dir} \
--train_dir ../save/train \
--test_dir ../save/test \
--phase train \
--epochs 200 \
--batch_size 12 \
--gpus 0 \
--test_gpu 0

Expected dataset structure

processed_dir
     ㄴ{dataset name}
         ㄴtrain: used for training
             ㄴcolor: directory which contains color images
             ㄴsketch: directory which contains sketch images
         ㄴval: used for validation
             ㄴcolor
             ㄴsketch
         ㄴtest: used for testing
             ㄴcolor
             ㄴsketch