YingqianWang

Excuse me. Have you ever trained the model before testing? If not, where did you get the trained model or weights?

Hello, the validation_bicubic.zip can be downloaded via Google Drive at https://drive.google.com/file/d/1gyaan54AwbAYLIIA1rly_wrLzdyQ7VAh/view?usp=sharing. You can refer to this example to format your submission, and the reported bugs will be fixed.

No, you needn't. For training data generation, just run the script ``GenerateTrainingData.m``. Then, training samples will be automatically saved as '.h5' files in the corresponding folder.

Thanks for your attention. At the very beginning, I noticed that images in the HCInew datasets are of low contrast. Then, I mannually adjusted their contrasts using a the Microsoft...

Thanks for your attention to our work. The pretrained models in this repo were trained using all the 6 datasets as mentioned in the ECCV paper. We also released the...

Thanks for your comments. I guess, that is probably due to the insuffient CPU memory of your PC. Since we did not crop the testing light field into sub light...

Thanks for your attention. We have checked our training codes and found that we had removed the initialization part by mistake during cleaning the codes. We have updated 'train.py' and...

For real-world SR, we directly applied SR methods to original HR images without any downsampling operation. Since the groundtruth HHR images are unavailable, we only compared the visual results of...

In our ECCV paper, only LFSSR was retrained since its official model was trained under the BD degradation (blur and downsampling), and other methods in this paper were directly tested...

Sorry for the late reply. We just followed the previous works to calculate the SSIM values. There are several versions of the SSIM evaluation pipeline. The calculation pipeline of SSIM...