codyshen0000 / itsrn

Code repo for "Implicit Transformer Network for Screen ContentImage Continuous Super-Resolution" (NeurIPS'21)

ITSRN

Code repo for "Implicit Transformer Network for Screen Content Image Continuous Super-Resolution"

itsrn's Issues

Can you provide the test configuration

test.py requires a configuration file in a specific format to run

Does the training use SCI1K dataset or SCI1K_Compression dataset?

Hi，I am confused about dataset and training .

In paper, you proposed two datasets, I can't tell which dataset you use in training, SCI1K or SCI1K_Compression?

On the basis of the original training set, I added some screen images (final dataset contain1500 images) and trained 2X model without jpeg preprocessing. I test the model using own testdataset(also without jpeg compression) and get bad subjective quality, and i test the your model on google drive, it's nice.

So I want to determine whether the problem is dataset preprocessing or adding new data。

Thanks for your good job, it's valueble for me。

Some pictures are missing under the SCI1K-Test folder

Hello, your work is great, I am very interested in your work, and now I am reproducing your work, I downloaded your data set, and found that in the SCI1K-Test directory, there are only 187 HR images, in the paper It is said that there are 200 pictures as a test set, can you release the complete test set?

Ask for code and dataset

Hi,
I just studied this impressive work. I am wondering will you release the code and the dataset?

Inquiry about Test code

Hi, I'm interested in your impressive work, ITSRN.

Could you please tell me your exact PSNR function and shaving factor for the dataset in your paper? (SCI1K, SCID, SIQAD)

Where is the code for "Implicit Position Encoding"？

Sorry to bother! But I can't find the section for the Implicit Position Encoding module.
Did I miss anything? Please inform.

Thank you for your assistance!

Test module not found

Hi, I am interested in this work and try to run your code. But the test code is absent and the train.py gets an error in line 39:

from test import eval_psnr
ImportError: cannot import name 'eval_psnr'

I hope that the complete code could be released soon.

SCI1K-compression JPEG compression quality factor

Hi, I have a question about SCI1K-compression dataset.

Your paper stated that the quality factor of JPEG compression is randomly selected from 75, 85, and 95.

However, in wrappers.py and image_folder_compress.py, the code for quality factor is commented out as "qf = random.randrange(10, 75)".

I think that to follow the statement in paper, "qf = random.randrange(75, 96, 10)" is right code.

What code did you use for training and testing on SCI1K-compression?