sRender

We provide Pytorch implementation for sRender (i.e. sketch render).

Reference:

Shang, Meimei, Fei Gao *, Xiang Li, Jingjie Zhu, and Lingna Dai. “Bridging Unpaired Facial Photos And Sketches By Line-drawings.” In ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 2010-2014. IEEE, 6-11 June 2021.

[paper@IEEE] [paper@arXiv] [project@github]

Pipeline

Sketch Reconstruction

left: croquis style; right: charcoal style (a)Sketches ; (b)Generated Line-drawings ; (c)Synthesized Sketches

Sketch Generation (from photos)

left: croquis style; right: charcoal style (a)Photos ; (b)Generated Line-drawings ; (c)Synthesized Sketches

Comparison

• Croquis sketches generated by our sRender and unpaired I2I translation methods

Train and test

• charcoal_style for systhesising charcoal style images,it contains sRender w/o Lstr for model without stroke_loss and sRender for model with stroke_loss correspodingly
• croquis_style for systhesising croquis style images,it contains sRenderPix2Pix, sRender w/o Lstr and sRender
• download dataset gray for charcol_style, binary for croquis_style
• download pretrain model and put them in ./checkpoints for test
• for model with stroke_loss, you should download stroke_model and specify model root in ./models/pix2pixHD_model for net_c.load_state_dict
• you can modify options/base_option to specify –dataroot, then run train.py or test.py

Dataset

• Gray line-drawings for charcol_style from GooleDrive
  • charcoal
  • CelebAmaskHQ
  • FFHQ
• Binary line-drawings for croquis_style from GooleDrive
  • croquis
  • CelebAmaskHQ
  • FFHQ

Pretrained models

• charcoal_style from GooleDrive
  • sRender w/o Lstr
  • sRender
• croquis_style from GooleDrive
  • sRenderPix2Pix
  • sRender w/o Lstr
  • sRender

Results

• Our synthesis result for croquis and charcoal style can be downloaded from Goole Drive

  Citation

bib

@inproceedings{shang2021bridging,
  title={Bridging Unpaired Facial Photos And Sketches By Line-drawings},
  author={Shang, Meimei and Gao, Fei and Li, Xiang and Zhu, Jingjie and Dai, Lingna},
  booktitle={ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  pages={2010--2014},
  year={2021},
  organization={IEEE}
}

Acknowledgments

• Our code is inspired by the AiSketcher repository and the NVIDIA/pix2pixHD repository.