Git Product home page Git Product logo

syn2real's People

Contributors

rajeevyasarla avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar  avatar

syn2real's Issues

About metrics of NIQE and BRISQUE

Thank you very much for your answer to my previous question about the real dataset.

Now I am confused by the non-reffereced metrics of NIQE and BRISQUE.
I test them with the code found online and find the results are quite different with those in your paper, even for the input real rainy images. The reason may be the uncorrect or different implementations of computing them.

So I am wondering if you can release the test code or provide an accesible link of the calculation of these two metrics.
Thanks again for your kindness.

About Real Rainy Images in Testing

Hello, Thanks for sharing this wonderful work.

I have a question about the testing data.
Are the images in the real-world test set from DDN-SIRR the same as those utilized in the training phase?

Thank you very much.

Data Problem

I think the data of 'norain' that you provided is not correct, could you please provide the corresponding dataset? Thanks.

How can i speed up the training period?

Hiiiiii, thanks for ur great job! i encounterd with the issue that the training time is too long(about 10h an epoch in deblur task) and i want to konw how can i speed up the training period. BTW, i want to konw how long about ur training time.
Looking forward for ur reply! thanks in advance!

A question for the inference phase

Hi,

Thanks for your contributions to the deraining research area. After going through the details of your work, I have a question about your algorithm. I notice that your real-world test dataset is the same as the real-world train dataset. My question is if I give you a new real-world rainy dataset, do you need to retrain the network? Or you can directly predict the results.

Looking forward to hearing from you.

Thanks.

Dataset for cross domain experiment.

Hi, I am curious about the experimental dataset.

You used 9,100 train images in DDN when you did real experiments.
Did you use the same 9,100 sheets for cross domain experiment? Or did you use 12,600 sheets as an official dataset?
I wonder if the Rain200H and Rain800 also used all of the training sets for cross domain experiment.

Test result problems about DDN_SIRR_with_GP

Hi, I train the model with
the 9100 pairs of images in the original dataset of Fu et al. [9] as the labeled image, the 147 images in real_input_split1.txt as the unlabeled image and the SIRR_test.txt as the val_filename;
but when I test the trained model on pictures in data/test/derain/SIRR_test, there are still some obvious white rain streaks on the SIRR_test, and the visual quality of test using retrained model (val_ssim is about 0.86) is much worser than using the DDN_SIRR_with_GP derain_best provided in the project(val_ssim is about 0.91).
Thank you very much!

Dataset Details (DDN-SIRR)

Hi, according to the paper, for the dataset DDN-SIRR, 9100 pairs of synthesized images and some real-world images are used as the training set, and another 20 pairs as testing set. I have some questions as follows:

  1. How many real-world images are in the training set? Are the 147 images in the path 'Syn2Real/data/train/derain/real_input/'?
  2. The original dataset of Fu et al. [9] consists of 9100 training pairs and 4900 testing pairs, did you use the 4900 pairs for testing?
  3. Could you share the data split file for the 9100 training pairs, since the ratio has been updated to 12600:1400 on their official websit.

Thank you very much!

Data Problem

I think the data of 'norain' that you provided is not correct, could you please provide the corresponding dataset? Thanks.

About other dataset

Can you provide the DID-Data、DDN-Data and SPA-Data on psnr/ssim result?

出现错误缺少文件

Traceback (most recent call last):
File "/home/zhao/lianxi/py1/S2/Syn2Real-master/test.py", line 77, in
os.mkdir('./{}_results/{}/'.format(category,exp_name))
FileNotFoundError: [Errno 2] No such file or directory: './derain_results/DDN_SIRR_withGP/'

The ground truth of unlabeled data.

Thanks for ur great job!
In the deblur task, i want to know what is the gt of the unlabeled sample during the unlabeled phase, i.e., in line 232th of "train.py".
Hope for your reply! Thanks in advance.

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.