DiffQRCoder: Diffusion-based Aesthetic QR Code Generation with Scanning Robustness Guided Iterative Refinement

Author: Jia-Wei Liao, Winston Wang, Tzu-Sian Wang, Li-Xuan Peng, Ju-Hsian Weng, Cheng-Fu Chou, Jun-Cheng Chen

IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2025

This repository implements a two-stage iterative refinement pipeline that leverages a pretrained ControlNet to generate aesthetic QR codes. Check out the project page here.

🔧 Setup

To set up the virtual environment and install the required packages, use the following commands:

virtualenv --python=python3.10 diffqrcoder
source diffqrcoder/bin/activate
pip install -r requirements.txt

🚀 Generating Aesthetic QR Code

To generate the aesthetic qrcode, run the following:

python run.py \
    --controlnet_ckpt <checkpoint of controlnet> \
    --pipe_ckpt <checkpoint of pipeline> \
    --qrcode_path <path of qrcode image> \
    --qrcode_module_size <qrcode module size> \
    --qrcode_padding <qrcode padding> \
    --neg_prompt <negative prompt> \
    --num_inference_steps <number of inference step> \
    --controlnet_conditioning_scale <controlnet conditioning scale> \
    -srg <scanning robust guidance scale> \
    -pg <perceptual guidance scale> \
    --srmpgd_num_iteration <number of srmpgd iterations> \
    --srmpgd_lr <srmpgd learning rate> \
    --device <device of running the code> \
    --output_folder <folder of generated image>

🔧 Unit Test

To execute the unit test, run the following:

bash scripts/run_test.sh

🎯 Citation

If you use this code, please cite the following:

@inproceedings{liao2024diffqrcoder,
  title     = {DiffQRCoder: Diffusion-based Aesthetic QR Code Generation with Scanning Robustness Guided Iterative Refinement},
  author    = {Jia-Wei Liao, Winston Wang, Tzu-Sian Wang, Li-Xuan Peng, Ju-Hsian Weng, Cheng-Fu Chou, Jun-Cheng Chen},
  booktitle = {IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
  year      = {2025},
}