Skip to content

PyTorch implementation of IEEE TPAMI 2021 paper: "Deep Spatial-angular Regularization for Light Field Imaging, Denoising, and Super-resolution".

Notifications You must be signed in to change notification settings

MantangGuo/DRLF

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 

Repository files navigation

DRLF

PyTorch implementation of IEEE TPAMI 2021 paper: "Deep Spatial-angular Regularization for Light Field Imaging, Denoising, and Super-resolution". The video demo is here

Requrements

  • Python 3.6.10
  • PyTorch 1.7.1
  • Matlab (for training/test data generation)

Compressive LF Reconstruction

1. Dataset

We provide MATLAB code for preparing the training and test data. Please first download light field datasets, and put them into corresponding folders in LFData.

2. Test

We provide the pre-trained models for tasks 1 -> 49, 2 -> 49, and 4 -> 49 on the Lytro dataset. Enter the LFCA folder and run:

Task 1 -> 49

python lfca_test.py --measurementNum 1

Task 2 -> 49

python lfca_test.py --measurementNum 2

Task 4 -> 49

python lfca_test.py --measurementNum 4

3. Train

Enter the LFCA folder and run:

python lfca_train.py

LF Denoising

1. Dataset

We provide MATLAB code for preparing the training and test data. Please first download light field datasets, and put them into corresponding folders in LFData. We used the same dataset, noise synthesis and preprocessing protocol as APA

2. Test

We provide the pre-trained models for adding zero-mean Gaussian noise with the standard variance varying in the range of 10, 20, and 50 on the Lytro dataset. Enter the LFDN folder and run:

Noise level 10

python lfdn_test.py --sigma 10

Noise level 20

python lfdn_test.py --sigma 20

Noise level 50

python lfdn_test.py --sigma 50

3. Train

Enter the LFCA folder and run:

python lfdn_train.py

LF Spatial SR

1. Dataset

We provide MATLAB code for preparing the training and test data. Please first download light field datasets, and put them into corresponding folders in LFData. We used the same dataset and protocol as those of Pseudo-4D to generate low-resolution LF images.

2. Test

We provide the pre-trained models for tasks 2x and 4x on the Lytro dataset. Enter the LFSSR folder and run:

Task 2x

python lfssr_test.py --scaleFactor 2

Task 4x

python lfssr_test.py --scaleFactor 4

3. Train

Enter the LFSSR folder and run:

python lfssr_train.py

About

PyTorch implementation of IEEE TPAMI 2021 paper: "Deep Spatial-angular Regularization for Light Field Imaging, Denoising, and Super-resolution".

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published