Awesome
FSPI-DM
Paper: High-resolution iterative reconstruction at extremely low sampling rate for Fourier single-pixel imaging via diffusion model
Authors: Xianlin Song, Xuan Liu, Zhouxu Luo, Jiaqing Dong, Wenhua Zhong, Guijun Wang, Binzhong He, Qiegen Liu
Optics Express 32 (3), 3138-3156, 2024
https://opg.optica.org/oe/fulltext.cfm?uri=oe-32-3-3138&id=545621
Date : Jan-9-2024
Version : 1.0
The code and the algorithm are for non-comercial use only.
Copyright 2024, Department of Electronic Information Engineering, Nanchang University.
Scheme of the system and the photographs of practical system.
<img src="Figures/1.png" alt="图片描述" width="100%" />Flow chart of high-resolution iterative reconstruction based on diffusion model.
<img src="Figures/2.png" alt="图片描述" width="100%" />The reconstruction results obtained by different methods for animal and coin under various sampling rates, as well as the corresponding ground truth and Fourier spectra.
<img src="Figures/3.png" alt="图片描述" width="100%" />Requirements and Dependencies
python==3.7.11
Pytorch==1.7.0
tensorflow==2.4.0
torchvision==0.8.0
tensorboard==2.7.0
scipy==1.7.3
numpy==1.19.5
ninja==1.10.2
matplotlib==3.5.1
jax==0.2.26
Checkpoints
We provide pretrained checkpoints of the dog. You can download pretrained models from [Baidu cloud] (https://pan.baidu.com/s/1IYIG5fQ_Ju_iRAbX455dSg) Extract the code (FSPI)
Dataset
- The data set used to train the model in this experiment comes from https://www.kaggle.com/datasets/unmoved/30k-cats-and-dogs-150x150-greyscale/data. We have extracted some uncontaminated images as training set, validation set and test set. Corresponds to "Training_set", "Validation_set" and "Test_set" in the warehouse
- The dog data used in the paper is located in the "Paper_data_dog" folder in the warehouse
Train:
- Replace the train_ds and eval_ds variables in the datasets.py file with the corresponding paths.
- Use the following command to train:
CUDA_VISIBLE_DEVICES=0 python main.py --config=aapm_sin_ncsnpp_gb.py --workdir=exp --mode=train --eval_folder=result
Test:
- Modify the ckpt_filename variable in A_PCsampling_demo.py to the corresponding checkpoint address
- Enter the file address of the low-frequency Fourier coefficients obtained after sampling into the y_k variable in the A_sampling.py file
- Use the following command to test:
CUDA_VISIBLE_DEVICES=0 python A_PCsampling_demo.py
Acknowledgement
The implementation is based on this repository: https://github.com/yang-song/score_sde_pytorch.
Other Related Projects
-
Diffusion Models for Medical Imaging <font size=5>[Paper]</font> <font size=5>[Code]</font> <font size=5>[PPT]</font>
-
Homotopic Gradients of Generative Density Priors for MR Image Reconstruction
<font size=5>[Paper]</font> <font size=5>[Code]</font> <font size=5>[PPT]</font> -
Multi-Channel and Multi-Model-Based Autoencoding Prior for Grayscale Image Restoration
<font size=5>[Paper]</font> <font size=5>[Code]</font> <font size=5>[Slide]</font> <font size=5>[数学图像联盟会议交流PPT]</font> -
Highly Undersampled Magnetic Resonance Imaging Reconstruction using Autoencoding Priors
<font size=5>[Paper]</font> <font size=5>[Code]</font> <font size=5>[Slide]</font> <font size=5>[数学图像联盟会议交流PPT]</font> -
High-dimensional Embedding Network Derived Prior for Compressive Sensing MRI Reconstruction
<font size=5>[Paper]</font> <font size=5>[Code]</font> -
Denoising Auto-encoding Priors in Undecimated Wavelet Domain for MR Image Reconstruction
<font size=5>[Paper]</font> <font size=5>[Paper]</font> <font size=5>[Code]</font> -
Learning Multi-Denoising Autoencoding Priors for Image Super-Resolution
<font size=5>[Paper]</font> <font size=5>[Code]</font> -
REDAEP: Robust and Enhanced Denoising Autoencoding Prior for Sparse-View CT Reconstruction
<font size=5>[Paper]</font> <font size=5>[Code]</font> <font size=5>[PPT]</font> <font size=5>[数学图像联盟会议交流PPT]</font> -
Iterative Reconstruction for Low-Dose CT using Deep Gradient Priors of Generative Model
<font size=5>[Paper]</font> <font size=5>[Code]</font> <font size=5>[PPT]</font> -
Universal Generative Modeling for Calibration-free Parallel MR Imaging
<font size=5>[Paper]</font> <font size=5>[Code]</font> <font size=5>[Poster]</font> -
Progressive Colorization via Interative Generative Models
<font size=5>[Paper]</font> <font size=5>[Code]</font> <font size=5>[PPT]</font> <font size=5>[数学图像联盟会议交流PPT]</font> -
Joint Intensity-Gradient Guided Generative Modeling for Colorization <font size=5>[Paper]</font> <font size=5>[Code]</font> <font size=5>[PPT]</font> <font size=5>[数学图像联盟会议交流PPT]</font>