Diversifying Detail and Appearance in Sketch-Based Face Image Synthesis

This code is our implementation of the following paper:

Takato Yoshikawa, Yuki Endo, Yoshihiro Kanamori: "Diversifying Detail and Appearance in Sketch-Based Face Image Synthesis" The Visual Computer (Proc. of Computer Graphics Internatinal 2022), 2022. [Project][PDF (28 MB)]

Prerequisites

Run the following code to install all pip packages.

pip install -r requirements.txt

Inference with our pre-trained models

Download our pre-trained models for the CelebA-HQ dataset and put them into the "pretrained_model" directory in the parent directory.
Download the zip file from "Human-Drawn Facial sketches" in DeepPS, unzip it, and put the "sketches" directory into the "data" directory in the parent directory.
Run test.py

cd src
python test.py

Training

Download the edge map dataset from Google Drive to the "data" directory.
Download the CelebA-HQ dataset and run resize_image.py to resize the image.

cd src
python resize_image.py --input path/to/CelebA-HQ/dataset --output ../data/CelebA-HQ256

Training the detail network H

python train.py \
--train_path ../data/CelebA-HQ256_DFE \
--edge_path ../data/CelebA-HQ256_HED \
--edgeSmooth \
--save_model_name network-H

Training the appearance network F

python train.py \
--train_path ../data/CelebA-HQ256 \
--edge_path ../data/CelebA-HQ256_DFE \
--weight_feat 0.0 \
--save_model_name network-F

Citation

Please cite our paper if you find the code useful:

@article{YoshikawaCGI22,
    author    = {Takato Yoshikawa and Yuki Endo and Yoshihiro Kanamori},
    title     = {Diversifying Detail and Appearance in Sketch-Based Face Image Synthesis},
    journal   = {The Visual Computer (Proc. of Computer Graphics Internatinal 2022)},
    volume    = {38},
    number    = {9},
    pages     = {3121-–3133},
    year      = {2022}
}

Acknowledgements

This code heavily borrows from the DeepPS repository.

tenten0727/Sketch-to-diverse-image