SketchColorization (Web)

Model Structure

Samples

GUI

---

Requirements

• torch==1.7.1
• torchvision==0.82
• numpy==1.19.1
• tensorboard==2.3.0
• tqdm==4.28.1
• opencv_python==4.4.0.46
• scipy==1.5.2
• Pillow==7.2.0
• scikit-learn==0.23.2
• fbs==0.9.0
• onnx==1.7.0
• onnxruntime==1.5.1
• PyQt5==5.15.1
• QDarkStyle==2.8.1

Dataset

• We crawled over 700,000 illustrations from shuushuu-image-board and used them for learning.

• We have filtered out noise such as extreme aspect ratio, black and white image, low / high key images and etc.

Training

• The learning sequence is 1. autoencoder, 2. draft, 3. colorization.

• set hyperparameters.yml, e.g. paths (image_path and line_path, logdir)

• Start learning after adjusting hyperparameters for each learning step

  • run 'python main.py -M {autoencoder | draft | colorization}'

Run APP with source code

• download pretrained onnx model SketchColorizationModel.onnx
• Copy model to "app/src/main/resources/base/SketchColorizationModel.onnx"
• cd app
• fbs run