Adaptive Deep PnP Algorithm for Video Snapshot Compressive Imaging (AdaptivePnP_SCI)

This repository contains the python code for the paper Wu, Z., Yang, C., Su, X., & Yuan, X. (2023). Adaptive Deep PnP Algorithm for Video Snapshot Compressive Imaging. International Journal of Computer Vision, https://doi.org/10.1007/s11263-023-01777-y..

Usage

Requirements

numpy==1.21.2
pytorch==1.8.1
scipy==1.7.1
scikit-image==0.18.1
h5py==2.10.0
tqdm==4.61.2
tensorboardx==2.2

Data

Please add simulation middle scale color data from PnP_SCI to ./dataset.

Our results are saved in OneDrive Link.

Test

Run ADMM_TV_Warm_Start_save.py to save the TV prior initialized results. [OPTION] Or just add the TV prior initialized results (_Admm_tv_xxx_bayer8.mat) from OneDrive Link to ./results/savedmat/.
Run twoStageADMM_Online_FFD_WARM.py or twoStageADMM_Online_FastDVD_WARM.py to test the algorithm after loading results initialized with TV prior as warn start.

Structure of directories

directory	description
`dataset`	data original
`packages` and `models`	algorithms pluged into PnP framework
`model_zoo`	pre-trained model data
`dataset`	data used for reconstruction (simulated or real data)
`results`	results of reconstruction (after reconstruction)
`utils`	utility functions

Citation

@article{wu_adaptive_2023,
 title = {Adaptive {Deep} {PnP} {Algorithm} for {Video} {Snapshot} {Compressive} {Imaging}},
 issn = {1573-1405},
 url = {https://doi.org/10.1007/s11263-023-01777-y},
 doi = {10.1007/s11263-023-01777-y},
 journal = {International Journal of Computer Vision},
 author = {Wu, Zongliang and Yang, Chengshuai and Su, Xiongfei and Yuan, Xin},
 month = mar,
 year = {2023},
}

xyvirtualgroup / AdaptivePnP_SCI