Quantify Free-Hand Sketch Quality

This repo contains the codebase of a series of research projects focused on quantifying human free-hand sketch quality

• Finding Badly Drawn Bunnies in CVPR2022
• Not That Bunny: An Annotation-Free Approach for Human Sketch Quality Assessment Under Review

Environment settings

-- Pytorch 1.7.1

Structure of this repo

dataloader.py -- load QuickDraw dataset

main.py -- main training and evaluation functions

models.py -- GACL implementations are here

data_utils.py -- sketch data processing functions

config.py -- experiment settings and hyper-parameters

Trained models

coming soon

Citation

If you use this code in your research, please kindly cite the following papers

@inproceedings{lancvpr,
  title={Finding Badly Drawn Bunnies},
  author={Yang, Lan and Pang, Kaiyue and Zhang, Honggang and Song, Yi-Zhe},
  booktitle={CVPR},
  year={2022},
}