BasicLFSR
BasicLFSR is an open-source and easy-to-use Light Field (LF) image Super-Ressolution (SR) toolbox based on PyTorch, including a collection of papers on LF image SR and a benchmark to comprehensively evaluate the performance of existing methods. We also provided simple pipelines to train/valid/test state-of-the-art methods to get started quickly, and you can transform your methods into the benchmark.
Note: This repository will be updated on a regular basis, and the pretrained models of existing methods will be open-sourced one after another. So stay tuned!
Methods
Methods | Paper | Repository |
---|---|---|
LFSSR | Light Field Spatial Super-Resolution Using Deep Efficient Spatial-Angular Separable Convolution. TIP2018 | spatialsr/ DeepLightFieldSSR |
resLF | Residual Networks for Light Field Image Super-Resolution. CVPR2019 | shuozh/resLF |
HDDRNet | High-Dimensional Dense Residual Convolutional Neural Network for Light Field Reconstruction. TPAMI2019 | monaen/ LightFieldReconstruction |
LF-InterNet | Spatial-Angular Interaction for Light Field Image Super-Resolution. ECCV2019 | YingqianWang/ LF-InterNet |
LFSSR-ATO | Light field spatial super-resolution via deep combinatorial geometry embedding and structural consistency regularization. CVPR2020 | jingjin25/ LFSSR-ATO |
LF-DFnet | Light field image super-resolution using deformable convolution. TIP2020 | YingqianWang/ LF-DFnet |
MEG-Net | End-to-End Light Field Spatial Super-Resolution Network using Multiple Epipolar Geometry. TIP2021 | shuozh/MEG-Net |
Datasets
We used the EPFL, HCInew, HCIold, INRIA and STFgantry datasets for both training and test. Please first download our datasets via Baidu Drive (key:7nzy) or OneDrive, and place the 5 datasets to the folder ./datasets/
.
-
After downloading, you should find following structure:
├──./datasets/ │ ├── EPFL │ │ ├── training │ │ │ ├── Bench_in_Paris.mat │ │ │ ├── Billboards.mat │ │ │ ├── ... │ │ ├── test │ │ │ ├── Bikes.mat │ │ │ ├── Books__Decoded.mat │ │ │ ├── ... │ ├── HCI_new │ ├── ...
-
Run
Generate_Data_for_Training.m
to generate training data. The generated data will be saved in./data_for_train/
(SR_5x5_2x, SR_5x5_4x). -
Run
Generate_Data_for_Test.m
to generate test data. The generated data will be saved in./data_for_test/
(SR_5x5_2x, SR_5x5_4x).
Benchmark
We benchmark several methods on above datasets, and PSNR and SSIM metrics are used for quantitative evaluation.
PSNR and SSIM values achieved by different methods for 2xSR:
Method | Scale | #Params. | EPFL | HCInew | HCIold | INRIA | STFgantry | Average |
---|---|---|---|---|---|---|---|---|
Bilinear | x2 | -- | 28.479949/0.918006 | 30.717944/0.919248 | 36.243278/0.970928 | 30.133901/0.945545 | 29.577468/0.931030 | 31.030508/0.936951 |
Bicubic | x2 | -- | 29.739509/0.937581 | 31.887011/0.935637 | 37.685776/0.978536 | 31.331483/0.957731 | 31.062631/0.949769 | 32.341282/0.951851 |
VDSR | x2 | |||||||
EDSR | x2 | 33.088922/0.962924 | 34.828374/0.959156 | 41.013989/0.987400 | 34.984982/0.976397 | 36.295865/0.981809 | ||
RCSN | x2 | |||||||
resLF | x2 | |||||||
LFSSR | x2 | 33.670594/0.974351 | 36.801555/0.974910 | 43.811050/0.993773 | 35.279443/0.983202 | 37.943969/0.989818 | ||
LF-ATO | x2 | 34.271635/0.975711 | 37.243620/0.976684 | 44.205264/0.994202 | 36.169943/0.984241 | 39.636445/0.992862 | ||
LF-InterNet | x2 | |||||||
LF-DFnet | x2 | |||||||
MEG-Net | x2 | |||||||
LFT | x2 |
PSNR and SSIM values achieved by different methods for 4xSR:
Method | Scale | #Params. | EPFL | HCInew | HCIold | INRIA | STFgantry | Average |
---|---|---|---|---|---|---|---|---|
Bilinear | x4 | -- | 24.567490/0.815793 | 27.084949/0.839677 | 31.688225/0.925630 | 26.226265/0.875682 | 25.203262/0.826105 | 26.954038/0.856577 |
Bicubic | x4 | -- | 25.264206/0.832389 | 27.714905/0.851661 | 32.576315/0.934428 | 26.951718/0.886740 | 26.087451/0.845230 | 27.718919/0.870090 |
VDSR | x4 | |||||||
EDSR | x4 | |||||||
RCSN | x4 | |||||||
resLF | x4 | |||||||
LFSSR | x4 | |||||||
LF-ATO | x4 | |||||||
LF-InterNet | x4 | |||||||
LF-DFnet | x4 | |||||||
MEG-Net | x4 | |||||||
LFT | x4 |
Train
- Run
train.py
to perform network training. Example for training [model_name] on 5x5 angular resolution for 2x/4x SR:$ python train.py --model_name [model_name] --angRes 5 --scale_factor 2 --batch_size 8 $ python train.py --model_name [model_name] --angRes 5 --scale_factor 4 --batch_size 4
- Checkpoints and Logs will be saved to
./log/
, and the./log/
has following structure:├──./log/ │ ├── SR_5x5_2x │ │ ├── [dataset_name] │ │ ├── [model_name] │ │ │ ├── [model_name]_log.txt │ │ │ ├── checkpoints │ │ │ │ ├── [model_name]_5x5_2x_epoch_01_model.pth │ │ │ │ ├── [model_name]_5x5_2x_epoch_02_model.pth │ │ │ │ ├── ... │ │ │ ├── results │ │ │ │ ├── VAL_epoch_01 │ │ │ │ ├── VAL_epoch_02 │ │ │ │ ├── ... │ │ ├── [other_model_name] │ │ ├── ... │ ├── SR_5x5_4x
Test
-
Run
test.py
to perform network inference. Example for test [model_name] on 5x5 angular resolution for 2x/4xSR:$ python test.py --model_name [model_name] --angRes 5 --scale_factor 2 $ python test.py --model_name [model_name] --angRes 5 --scale_factor 4
-
The PSNR and SSIM values of each dataset will be saved to
./log/
, and the./log/
is following structure:├──./log/ │ ├── SR_5x5_2x │ │ ├── [dataset_name] │ │ ├── [model_name] │ │ │ ├── [model_name]_log.txt │ │ │ ├── checkpoints │ │ │ │ ├── ... │ │ │ ├── results │ │ │ │ ├── Test │ │ │ │ │ ├── evaluation.xls │ │ │ │ │ ├── [dataset_1_name] │ │ │ │ │ │ ├── [scene_1_name] │ │ │ │ │ │ │ ├── [scene_1_name]_CenterView.bmp │ │ │ │ │ │ │ ├── [scene_1_name]_SAI.bmp │ │ │ │ │ │ │ ├── views │ │ │ │ │ │ │ │ ├── [scene_1_name]_0_0.bmp │ │ │ │ │ │ │ │ ├── [scene_1_name]_0_1.bmp │ │ │ │ │ │ │ │ ├── ... │ │ │ │ │ │ │ │ ├── [scene_1_name]_4_4.bmp │ │ │ │ │ │ ├── [scene_2_name] │ │ │ │ │ │ ├── ... │ │ │ │ │ ├── [dataset_2_name] │ │ │ │ │ ├── ... │ │ │ │ ├── VAL_epoch_01 │ │ │ │ ├── ... │ │ ├── [other_model_name] │ │ ├── ... │ ├── SR_5x5_4x
Recources
We provide some original super-resolved images and useful resources to facilitate researchers to reproduce the above results.
Other Recources
- YapengTian/Single-Image-Super-Resolution
- LoSealL/VideoSuperResolution
- ChaofWang/Awesome-Super-Resolution
- ptkin/Awesome-Super-Resolution
- lightfield-analysis/resources
- Joechann0831/LFSRBenchmark
- YingqianWang/LF-Image-SR
Contact
Any question regarding this work can be addressed to [email protected].