miccunifi / tape Goto Github PK

How can this be used in a gui interface like comfy ui? I am trying to route it through multiple models.

Are there plans to release a docker image or or some kind of GUI to test this out?

I would love to use in a video workflow. For example using this with depthai-gui

Thank you for your hard work!

Hi, This work is so wonderful. I am interested in VHS video and try to restore my own VHS tapes. I am wondering when would the code and the pretrained model be released?

Thanks

I tried this, but i get always an error, that frames are failed to read. I tried different videos and i tried .mkv and .mp4 but it's always the same. I also tried to manually define the checkpoint.pth because i thought maybe it is wrong. But i still get the same error.

I got this error:

(TAPE) userr@USERRNAME:~/TAPE$  python real_world_inference.py --input-path /home/userr/Videos/vhstest/VHS-Kamera-Test.mp4 --output-path /home/userr/Videos/vhstest/output/VHS-Restauration1.mp4
Extracting frames from the video...
 99%|███████████████████████████████████████████████████████████████████████████████▌| 160/161 [00:01<00:00, 111.88it/s]
Traceback (most recent call last):
  File "/home/userr/TAPE/real_world_inference.py", line 235, in <module>
    real_world_test(args)
  File "/home/userr/TAPE/real_world_inference.py", line 52, in real_world_test
    raise Exception("Failed to read frame from video")
Exception: Failed to read frame from video

First it looks like it would work, but then it always stops on the second last frame.

Edit: It does run now with original (uncut) video. When i cut it to a short one with avidemux or virtualdub2 it does not work anymore. Dont know why. But it runs now.

Hello, would it be possible to share your Adobe After Effects project files (maybe containing a small video that we can replace with our own) or explain the step-by-step procedure how you made the synthetic artifacts? I'm interested in restoring some historical cartoon videos, so I need a different kind of synthetic dataset.

I got the error below in the picture. I dont know what is wrong. It does not extract frames, but i dont know why. I dont know what these errors means. Do you have any idea ?

Congratulations for this work, the result is incredible, the link to the pre-trained model does not work?

Im currently facing this issue

I tried to run the script 2 times but I got the same error.

I have a RTX 3060ti 8gb no other apps running at same time fully dedicated to this

Hey there thanks for this, I'm trying to revive old videos from my grand mother, thanks for this!

I'm trying to run this script

C:\Users\StudioPC\Documents\TAPE\TAPE> python real_world_inference.py --input-path "C:\Users\StudioPC\Videos\JVC-VHSC" --output-path "C:\Users\StudioPC\Videos\JVC-VHSC\Output"

at first try it started to classify images.

but then I'm getting this error

Then I erased the output folder and tried again the first script and i got this

all the videos on the Input folder are "mp4"

Am I missing some args?

Thanks in advance!

Your research is really stunning. Thank you for aiming at analog video as this is an area that people forget exists. Are you and your team considering mixing your work with current methods of restore/upscaling in addition to the work of correcting analog errors/timing errors? Examples of this would be the more recent "HAT: Hybrid Attention Transformer for Image Restoration" or BasicVSR++. There are in depth research methods available as well as code examples.

HAT: https://github.com/xpixelgroup/hat
BasicVSR++ : https://ckkelvinchan.github.io/projects/BasicVSR++/

It could be used to restore more analog details once your method has finished grabbing what it can.

I can't wait to see what you all come up with next. Truly Exciting!
Thank you for sharing this with the world!