Yuelin Zhang, Pengyu Zheng, Wanquan Yan, Chengyu Fang, Shing Shin Cheng; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 11125-11136

paper / supplementary material

We release CataBlur dataset. CataBlur is a new surgical microscope defocus blur dataset, including 1208 defocus images collected from 5 cataract surgeries, for evaluation on real surgical defocus blur.

It can be downloaded from Google Drive.

Other open-source datasets can be downloaded from their original websites.

All downloaded datasets should be placed under the Datasets folder.

conda create -n MPT python=3.11
conda activate MPT
# If necessary, Replace pytorch-cuda=? with the compatible version of your GPU driver.
conda install pytorch==2.0.0 torchvision==0.15.1 torchaudio==2.0.1 pytorch-cuda=11.8 -c pytorch -c nvidia
pip install -r requirements.txt

To train the model with multiple GPUs, use the script.

sh train_dist.sh

To train the model with a single GPU, call the python script directly.

python run.py -net MPT -config config_MPT -data DATASETS_THAT_YOU_WANT_TO_USE -train True

All model weights can be downloaded at here. Put the weight in Weights directory.

python run_inference.py --data /path/to/your/images --pretrained /path/to/your/weights

python run.py -net MPT -config config_MPT -data DATASETS_THAT_YOU_WANT_TO_USE \
        -train False -ckpt_name /path/to/your/weights

If you find our work helpful, please consider citing our paper:

@inproceedings{zhang2024unified,
  title={A unified framework for microscopy defocus deblur with multi-pyramid transformer and contrastive learning},
  author={Zhang, Yuelin and Zheng, Pengyu and Yan, Wanquan and Fang, Chengyu and Cheng, Shing Shin},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={11125--11136},
  year={2024}
}