Deinterlacing DVD NTSC 480i to 480p

[DruidCtba]

Attached an example of using conventional (no NVIDIA card or ml stuff aside from NNEDI3 + can be done with Avisynth&Vapoursynth) filtering. (script: https://pastebin.com/vKLn0cRJ)

Cu Selur

That's gone too far, it's clearly degraded the quality. (haloes, no real detail, etc)
Perhaps 720p would be bettter - unless the 1080p takes priority based on the resolution alone?

Apparently yes, my friend @davexnet, as the degradation is visible, you could even notice some tremors, anyway, in the previous case my friend @Selur used specific commands for the NVIDIA GPU, if I'm not mistaken, and also the CodeFormer software, which adds considerable value to the faces, which for me made all the difference, anyway, apart from the problem with the musician's nose, generated by Topaz's DIONE TV, the rest of the show performed reasonably well for 1080p, which is the definition that look.

Att.

Druid®.

I agree, I thought Selurs previous version was very good. What is the casue of Topaz causing such damage to the musicians nose?
Is it an aggressive setting somewhere?

Look my friend @davexnet, I have some guesses (but if you notice, it's not all musicians, I only detected one):

1. The musician was not wearing makeup, which makes his skin glow.
2. The light artifacts on the nose, on the side, appear in the shadow of it, if you notice it in the original.
3. Perhaps the combination of the two mentioned above made the AI algorithm get lost and want to reproduce something that didn't exist.

Att.

Druid®.

[DruidCtba]

@davenext: used TemporalDegrain2, using SMDegrain or simply adjusting the strenght of the filtering would probably be less destructive.
@DruidCtba: did a quick testing on a Ryzen 9 9850 with an NVIDIA 4080 and got around ~2.5fps, so my system it would probably take a similar time that it took you on yours to run Topaz.

Friend @Selur it is a reasonable time to obtain better quality than software costing almost US$400, so it is worth every penny to produce a final product with all your expertise.

I will be very happy when I have a similar configuration to be able to work on my Uscaling here in Brazil .

It's a shame that some software only uses NVIDIA hardware to operate, anyway, I even have an NVIDIA media player, the SHIELD TV 4K, but from 2017, and I know that there are more current models of this media player that have built-in upscaling that generates excellent output quality, in real time, without you having to resort to manual upscaling.

Att.

Druid®.

[DruidCtba]

For anyone interested, I released a 4K version of Yanni's show, on rutracker (https://rutracker.org/forum/viewtopic.php?t=6058539).

As I mentioned previously, the 4K version did not generate the artifacts on the musician's nose, and I believe the result was satisfactory.

I sincerely thank you all, especially my friend @Selur, for your help here on the forum, may God always bless you for your altruism and selflessness in helping everyone here .

Att.

Druid®.

P.S. I always prefer the FullHD versions, because my DUNE SOLO LITE 4K media player doesn't play these settings well in my setup, but as I also have an NVIDIA SHIELD TV 4K, I run it, smooth and without choking, but the DUNE is still mine most used media player.

[DruidCtba]

Friend @Selur, is there any way I can use CodeFormer for video?

I don't know how to use it, I saw on the website that it only exists for Images, or am I mistaken?

As you did there, I wanted to do a test here, using CodeFormer for video and then insert the result with the faces corrected for Topaz Video AI v.4.0.5.

Att.

Druid®.

[Selur]

I used CodeFormer through vs-codeformer (https://github.com/HolyWu/vs-codeformer) it still works on a per-frame basis, but on a whole video.

[DruidCtba]

I used CodeFormer through vs-codeformer (https://github.com/HolyWu/vs-codeformer) it still works on a per-frame basis, but on a whole video.

Thank you very much good friend .

Att.

Druid®.

[DruidCtba]

I used CodeFormer through vs-codeformer (https://github.com/HolyWu/vs-codeformer) it still works on a per-frame basis, but on a whole video.

Thank you very much good friend .

Att.

Druid®.

So my friend @Selur, I confess that I just want to get the original 480i video, and use CodeFormer to arrange the faces, because structures and backgrounds, Topaz Video AI v.4.0.5, does a good job when Upscaling to 1080p and/or 4K, but I confess that I didn't understand how to use it, there are a lot of commands, and I got completely lost here; because my idea is to better the faces with Codeformer without touching the background and then applying this result in Topaz, is this possible?

I don't know what the output result of CodeFormer is for a video, I saw that it has the option on the website, where you upload a photo, there is something like Rescaling_Factor (up to 4), which seems to double the size of the image, and so I ask, Would the result for video be an Interlaced video or a Progressive video?

I don't want to process the video itself, just improve the faces and then do an Upscaling in Topaz, is this possible? Did you understand my question?

Codeformer doesn't have a user guide explaining how to use it, I mean, how could a layman use it and then I got really lost :/.

Att.

Druid®.

P.S. OK, user guide, simple, here - https://github.com/sczhou/CodeFormer.

[DruidCtba]

Why am I getting this error, is it because I don't have an NVIDIA card to handle CUDA?

C:\Files\ffmpeg-2023-11-27-git-0ea9e26636-full_build\CodeFormer-master>python inference_codeformer.py --bg_upsampler realesrgan --face_upsample -w 1.0 --input_path "C:\Files\ffmpeg-2023-11-27-git-0ea9e26636-full_build\CodeFormerVideo\Yanni - Tribute (1997 DVD-9 480i Original)_0001.mp4"
Traceback (most recent call last):
File "C:\Files\ffmpeg-2023-11-27-git-0ea9e26636-full_build\CodeFormer-master\inference_codeformer.py", line 5, in <module>
import torch
File "C:\Users\FILE X\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\__init__.py", line 1429, in <module>
from torch import optim as optim
File "C:\Users\FILE X\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\optim\__init__.py", line 8, in <module>
from .adadelta import Adadelta
File "C:\Users\FILE X\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\optim\adadelta.py", line 4, in <module>
from .optimizer import (Optimizer, _use_grad_for_differentiable, _default_to_fused_or_foreach,
File "C:\Users\FILE X\AppData\Local\Programs\Python\Python311\Lib\site-packages\torch\optim\optimizer.py", line 23, in <module>
from typing_extensions import ParamSpec, Self, TypeAlias
ImportError: cannot import name 'Self' from 'typing_extensions' (C:\Users\FILE X\AppData\Local\Programs\Python\Python311\Lib\site-packages\typing_extensions.py)

C:\Files\ffmpeg-2023-11-27-git-0ea9e26636-full_build\CodeFormer-master>

Att.

Druid®.

[Selur]

I don't have an NVIDIA card to handle CUDA?

afaik. codeformer requires an NVIDIA card.

[DruidCtba]

I don't have an NVIDIA card to handle CUDA?

afaik. codeformer requires an NVIDIA card.

Att.

Druid®.

[DruidCtba]

I don't have an NVIDIA card to handle CUDA?

afaik. codeformer requires an NVIDIA card.

Att.

Druid®.

[lordsmurf]

Topaz AI crap.
And b.t.w. why always jump onto the bandwagon of deinterlacing and upscaling? Too many good video is just getting spoiled....

Repeated for emphasis.

Topaz = crap
upscale = generally done for no good reason

[DruidCtba]

Topaz AI crap.
And b.t.w. why always jump onto the bandwagon of deinterlacing and upscaling? Too many good video is just getting spoiled....

Repeated for emphasis.

Topaz = crap
upscale = generally done for no good reason

Well, as we live in a democracy, "I think", I will disagree with some opinions, although I respect them and defend each point of view, I prefer to always keep the original, but I have no doubt that I prefer a thousand times to watch this show with Upscaling by Topaz, on my 50" Samsung TU8000 TV, than on the original DVD in 480i, but hey, this is my opinion, and since I keep the original, I don't see myself ruining anything.

Anyway, dead king, deposed king, this is democracy, that's why many hate it, but I, on the other hand, love freedom, and I will always defend it, if everything were the same, as some think they want, life would be much more boring than who are wanting to make it.

Att.

Druid®.