Temporal Filter bug in Davinci Resolve

resolve technical issues related to use of Neat Video
Post Reply
misha
Posts: 5
Joined: Tue Sep 16, 2014 8:11 am

Temporal Filter bug in Davinci Resolve

Post by misha »

I'm using NV3 and upgraded to NV4 for Davinci Resolve.
NV3 has perfect accuracy of temporal filter (especially with radius=5)
NV4 does blurry spots, lost details and result look far worse than preview.

Sorry for big picture.
Image

My Cuda 5.2 videocard is not supported in NV3 and nextgen version works inaccuracy.
:(
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

1. Make sure you build the noise profile correctly, that is most important.

2. Check preview inside Resolve's window, not in Neat Video, because Resolve does not provide all input frames required to prepare accurate preview (specifically, the temporal filter cannot work without those missing input frames). In Resolve preview, the filter is able to receive all input frames and prepare correctly processed preview.

Thank you,
Vlad
misha
Posts: 5
Joined: Tue Sep 16, 2014 8:11 am

Post by misha »

NVTeam wrote:1. Make sure you build the noise profile correctly, that is most important.

2. Check preview inside Resolve's window, not in Neat Video, because Resolve does not provide all input frames required to prepare accurate preview (specifically, the temporal filter cannot work without those missing input frames). In Resolve preview, the filter is able to receive all input frames and prepare correctly processed preview.

Thank you,
Vlad
Noise profile is correct. I'm using NV for years and purchased 3 copies.

NV3 temporal filter works perfect without losing details but CUDA 5.2 unsupported.
NV4 is unacceptable because of that blurry areas with losing details.

In another topic you say you plan to include CUDA 5.2 support in the next regular update for NV3. Is it cancelled?
https://www.neatvideo.com/nvforum/viewtopic.php?t=1100

thanks
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

Please prepare a reproducible test case - a small project showing both NV3 and NV4 on a short test clip. We will run it on our machines and check what exactly is causing the problem. It most likely can be avoided or fixed.
Please send that test case to support [at] neatvideo.com

If you want to try v3 with 5.2 support please also send a request to support [at] neatvideo.com

Thank you,
Vlad
misha
Posts: 5
Joined: Tue Sep 16, 2014 8:11 am

Post by misha »

NVTeam wrote:Please prepare a reproducible test case - a small project showing both NV3 and NV4 on a short test clip. We will run it on our machines and check what exactly is causing the problem. It most likely can be avoided or fixed.
Please send that test case to support [at] neatvideo.com

If you want to try v3 with 5.2 support please also send a request to support [at] neatvideo.com

Thank you,
Vlad
thanks!
downloaded new version of NV3. It works!

I'll prepare for you test on next week.
jpsdr
Posts: 221
Joined: Mon Aug 11, 2008 7:33 am

Post by jpsdr »

I'm interested to see the results of this investigation.
misha
Posts: 5
Joined: Tue Sep 16, 2014 8:11 am

Post by misha »

Sorry for delay. I will do it soon. Too much work (almost done).
misha
Posts: 5
Joined: Tue Sep 16, 2014 8:11 am

Post by misha »

I'm here!

XAVC from A7S
Davinci Resolve 11.3
https://yadi.sk/d/lSdk0di5iQ4Wq

No effect
Image

NV4 no temporal
Image

NV4 temporal=5
Image

NV3 temporal=5 (same noise mask)
Image

NV3 is cleaner and details much better

I can do more samples if you need.
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

Thank you for the test case.

To compare results like this they have to be prepared based on the same noise profiles and with the same filter settings. We will download the files and check if these conditions are satisfied in the test project.

Thank you,
Vlad
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

We have opened the test project and found that:
- there supposed to be two clips in the timeline but only one clip is actually present for testing;

- there are no effects added to these clips, so there are no any instances of NV4 or NV3;

- so there is no way for us to directly see/reproduce the results you posted yet NV4 or NV3 are not in the test project yet.


Could you please add instance(s) of NV to the clip or clips and then configure them the way you did to get the screenshots you posted?

Then we would be able to directly reproduce your results on our machines. Without that, we do not know what kind of noise profiles you built, what filter settings you used, etc.

Thank you,
Vlad
Post Reply