MedVSR: Medical Video Super-Resolution with Cross State-Space Propagation
This model was presented in the paper MedVSR: Medical Video Super-Resolution with Cross State-Space Propagation. The official code repository can be found at: https://github.com/CUHK-AIM-Group/MedVSR.
Overview
MedVSR is a tailored model for medical VSR. It first employs Cross State-Space Propagation (CSSP) to address the imprecise alignment by projecting distant frames as control matrices within state-space models, enabling the selective propagation of consistent and informative features to neighboring frames for effective alignment. It also features an Inner State-Space Reconstruction (ISSR) module that enhances tissue structures and reduces artifacts with joint long-range spatial feature learning and large-kernel short-range information aggregation.
Installation
Clone this repository:
git clone https://github.com/CUHK-AIM-Group/MedVSR
cd MedVSR
conda create -n MedVSR python==3.9
conda activate MedVSR
pip install torch==2.1.1+cu121 torchvision==0.16.1+cu121 --extra-index-url https://download.pytorch.org/whl/cu121
pip install -r requirements.txt
pip install -e causal_conv1d>=1.1.0
pip install -e mamba-1p1p1
Dataset preparation
For the preprocessed HyperKvasir, LDPolyp, and EndoVis18, please download from huggingface link. Modify L14-16 and L39-40 to the extracted HyperKvasir training and validation folders.
Test the model
Download our pretrained model at here.
python test_model.py -opt ./options/medvsr_train.yml --weight <PATH_TO_PRETRAINED_MEDVSR>
Training
bash dist_train.sh 2 options/medvsr_train.yml 25623
Citation
@inproceedings{liu2025medvsr,
title = {MedVSR: Medical Video Super-Resolution with Cross State-Space Propagation},
author = {Liu, Xinyu and Sun, Guolei and Wang, Cheng and Yuan, Yixuan and Konukoglu, Ender},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
year = {2025}
}
Acknowledgement
We sincerely thank the authors and contributors of the following projects for their awesome codebases, which have greatly benefited our work:
Contact
Please contact [email protected] or open an issue.