A question about the code for generating distorted images. about pipal-dataset HOT 5 CLOSED

Hi, is it possible to release the code for generating distorted images? This code may be important for further IQA research, such as generating more distorted images for data augmentation.

Thanks for paying attention to our research.

Currently, we do not have such an integrated codebase to generate all distorted images because of the traditional super-resolution method and version consistency (especially TensorFlow). We know such a tool that can generate all distorted images will benefit the whole community. However, it is a huge workload to unify every Image Processing method to one platform. We must consider that most old models (from 2015 to 2018) are hard to reproduce due to version updates.

As you can see, this problem is also a challenge in the IQA field. There isn't a convenient and open-source tool led by a large research group or research company.

If you have any other questions, please let me know.

from pipal-dataset.

Thank you for your reply. Didn't you save the source code when developing this project?

from pipal-dataset.

Can you send me a copy of the source code if you have saved it. I can do the rest of the work myself. Thank you very much.

from pipal-dataset.

Can you send me a copy of the source code if you have saved it. I can do the rest of the work myself. Thank you very much.

I suggest you collect them yourself according to the algorithms listed in our Journal version (on ArXiv). We adapted and slightly modified those official codes to our local environment. Therefore, it might cost you more time to adjust our code back to the original one. If you are familiar with TensorFlow / PyTorch / Matlab / Linux, I think 2 weeks are enough for you to construct a proper environment for most of the algorithms based on their README.

There are parts of distortions (like spatial warping) created by ourselves. I can share those with you. I will attach a Google Link to this issue ASAP.

from pipal-dataset.

Thank you for your patience in replying, and for your suggestions and sharing.

from pipal-dataset.