search

ryanxingql / stdf-pytorch

Implementation of "Spatio-Temporal Deformable Convolution for Compressed Video Quality Enhancement" (AAAI'20).
https://www.aiide.org/ojs/index.php/AAAI/article/view/6697
Apache License 2.0
155 stars 20 forks source link

Dropbox dataset not available #12

Closed gulzainali98 closed 2 years ago

gulzainali98 commented 2 years ago

Dataset in dropbox is not available. However, I am not sure how to download the baidu pan dataset directly to my server.

gulzainali98 commented 2 years ago

It's necessary for me because i have to download the data directly to my cluster

ryanxingql commented 2 years ago

Hi gulzainali98, sorry for my late response. I will fix it in one day.

gulzainali98 commented 2 years ago

Thank you for responding.

I needed the dataset for a project I am working on..... would be great if this can get resolved as soon as possible :)

Thank you very much. Or maybe you can provide any other option. Maybe google drive :/

ryanxingql commented 2 years ago

Sorry for my late response. The Dropbox subscription is out-of-date now and I will fix it in one day. I will also contact you as soon as I finish it. Sorry again.

ryanxingql commented 2 years ago

Hi gulzainali98! A specialized team from Dropbox is helping us to fix it. It may take 24 hours. Sorry for your inconvenience. As soon as it is okay, I will contact you.

gulzainali98 commented 2 years ago

hello, Okay looking forward to your response :)

ryanxingql commented 2 years ago

Hi, The former Dropbox link is broken. I create a new link and I am still uploading files to this link, which may take about 48 hours. Sorry for your inconvenience. When I finish uploading, I will inform you.

ryanxingql commented 2 years ago

Hi, you can download the data-set from the new Dropbox link now!

gulzainali98 commented 2 years ago

Hey Sorry for late response. I can download the dataset. Thank you very much. However, do you know if the video encoder can be used to simply compress set of images instead of a video?

ryanxingql commented 2 years ago

You can check here: https://github.com/RyanXingQL/PowerQE#2-image-data I suggest using the BPG encoder.

gulzainali98 commented 2 years ago

BPG will result same compression artifacts as HVEC? Can I replace HVEC with BPG and train your network on this data?

ryanxingql commented 2 years ago

BPG adopts the All Intra (AI) mode of HEVC. In MFQE and STDF papers, LDP mode is considered. The AI and LDP modes are different, so as the resulting compression artifacts.

gulzainali98 commented 2 years ago

I can create a video out of images but when can't seem to extract frames from the compressed video. I use opencv I420 to create a yuv video from the frames and then use encoder provided to compress it and then i want to extract the frames from this compressed video.

gulzainali98 commented 2 years ago

rgb frames*

gulzainali98 commented 2 years ago

i was able to extract the frames. Do you use I420 yuv videos in MFQE dataset?

ryanxingql commented 2 years ago

YUV 420P format.

gulzainali98 commented 2 years ago

Any idea how to use HEVC on custom width and height? I am getting following error Resulting coded frame width must be a multiple of the minimum CU size

ryanxingql commented 2 years ago

You can first pad the input frame, and make sure that the new height and width are a multiple of 16. After compression, you can crop the input frame back with the original resolution.

gulzainali98 commented 2 years ago

yes. Thank you very much. perfect solution (y)

gulzainali98 commented 2 years ago

still HEVC not working my frame size in videos is (48, 64, 3).... could it be that the frame size is too small?

ryanxingql commented 2 years ago

You can try 64 instead of 16.