Miscellaneous Forums > Misc. Tools

Video Restoration with Enhanced Deformable Convolutional Networks

(1/2) > >>

Fraggoso:
I didn't had the time to test this yet, but it seems promising for uprezsing Videos. It's also from some of the main contributors of ESRGAN.
https://github.com/xinntao/EDVR

Someone already had time to test this with Final Fantasy Movies?

satsuki:
Look great tool but can't get it works on windows (can't build some depencencies)
Will install a linux distrib in some days to test it

satsuki:
Try to use it but i issue a bug (open an issue in git but posting it here too ^^)

I'm trying to use the test_Vid4_REDS4_with_GT.py script with a foder containing 98 png files (named from 00000001.png to 00000099.png).
The png resolution is 320*240.

when i use the script i got this error:


--- Code: ---  File "test_Vid4_REDS4_with_GT.py", line 276, in <module>
    main()
  File "test_Vid4_REDS4_with_GT.py", line 217, in main
    GT = np.copy(img_GT_l[img_idx])
IndexError: list index out of range
--- End code ---

And when i look at result i can see it have processed the first png file but not the others

Thanks for helping ^^

Manakaiser:
After a quick check, Its possibly due to forced psnr calculation which requires a GT for each image.

Try creating a GT image for each frame in GT folder (just use a duplicate of the LR Frames, since its irrelevant which result the PSNR calculation gives). You might have to make the image copies in the GT folder the same resolution as your resuslts are expected to be, supposedly *4 of the LR provided you try to upscale by a factor of 4.



Let me know if it works.

If after testing it Works: You can try to avoid this step in the future by trying to remove  all lines from 215 to 273. And probably 169 to 179. But that might not work, just a suggestion if the initial advice resolves the problem for you.

satsuki:
I 've done it, it's working  but i get artifact in some pictures, so no usable so far :


Seems related to model training.
I don't have time to try a model of my own but on picture without artifact seems promising as a prefilter to use before esrgan

Navigation

[0] Message Index

[#] Next page