Livepano layer mask visible in chrome

[benji33]

Re: Livepano layer mask visible in chrome

by benji33 » Wed Oct 25, 2017 12:51 pm

@ungeherr Here is my observation with your project sample without any change on it.

My configuration : Windows 10 64bits + NVIDIA Geforce GT 640 2Mo DDR3

- Chrome with hardware acceleration (62.0.3202.62 64bits) : Mask is visible
- Chrome without hardware acceleration [changed into advanced parameters] (62.0.3202.62 64bits) : OK
- Firefox (56.0.1 64 bits) with hardware acceleration : Mask is visible
- Firefox (56.0.1 64 bits) without hardware acceleration [changed into about:config] : Mask is visible
- Firefox (56.0.1 32 bits) with hardware acceleration : OK
- MS Edge (Edge HTML 15.15063) : OK
- IE 11 (11.674.15063.0) : OK

This is the first step of my verification.

I cannot see any real difference on Chrome and Firefox 64bits when using the hitfim video directly into the generated project.

This is the second step of my verification.

I don't have the images resources linked to the .pano to be able to build the Livepano into the project on my side.

EDIT: I've tested with an other livepano, thanks to R.Gramolini and issue is present.

This is the third step of my verification.

I've added a new test with a Win 10 64bits with an AMD R9 200 series.

It's OK under Chrome 64bits and Firefox 64bits.
So all is convergent to an issue linked to NVIDIA graphics cards when used for hardware accelerated video decoding.
It's also one of the reason that many softwares make the video copy through CPU to avoid colorimetry issues.
Here is one of the chromium issues : https://bugs.chromium.org/p/chromium/is ... ?id=728670
Here is a NVIDIA issue linked too : https://forums.geforce.com/default/topi ... e-balance/

If it's the case there is nothing to do with Panotour to solve this hardware/drivers issue!
But a flag can be raised on Chromium and Mozilla for that.
Chrome Bug Report : http://dev.chromium.org/for-testers/bug ... guidelines
Firefox Bug Report : https://bugzilla.mozilla.org/page.cgi?id=etiquette.html

This is the fourth step of my verification.

[ungeherr]

Re: Livepano layer mask visible in chrome

by ungeherr » Wed Oct 25, 2017 9:04 pm

Hi Benjamin,

I'm sorry. My instructions were unclear. Or actually, I forgot to include them.

The "project file" RAR included the following
* The internet files with the original untouched livepano (not processed though hitfilm)
* .kpt project file
* .pano file
* video source file
* video png for the .pano file
* The equirectangular panorama

Everything you needed to remove and create the livepano again if needed.
I didn't include the RAW files to save space. Didn't see why you need them since since you already had the panorama

So what you need to do now to really see a difference is to take the "backdoor_4_vid_1_livepano_high.rar" and unrar it in the spots folder to replace the original livepano.

I've viewed the original VS hitfilm version of the livepano on both AMD and Nvidia based PCs and hitfilm works (makes a positive difference) for both of them.

[benji33]

Re: Livepano layer mask visible in chrome

by benji33 » Wed Oct 25, 2017 9:26 pm

During my test I can't see any difference with my PC with NVIDIA card.
I'll have a new test on NVIDIA and AMD graphics cards tomorrow.

The img sources files are required by the .pano to check the snapshot position. I agree it will be better if the .pano only was required.

[ungeherr]

Re: Livepano layer mask visible in chrome

by ungeherr » Wed Oct 25, 2017 9:29 pm

You see no difference between these examples?

http://tour.swedvr.se/thefan_unprocessed/thefan.html
http://tour.swedvr.se/thefan_through_hi ... hefan.html

http://tour.swedvr.se/livepanonightclub/index.html
http://tour.swedvr.se/hitfilmlivepanoni ... index.html

[ungeherr]

Re: Livepano layer mask visible in chrome

by ungeherr » Wed Oct 25, 2017 9:39 pm

Here are the RAWs
https://drive.google.com/open?id=0B8N9q ... E9qZmdITUk

[ERR]

Re: Livepano layer mask visible in chrome

by ERR » Thu Oct 26, 2017 12:32 am

I can see a marked difference in the nightclub and slightly less with the fans but the difference is measurable. Both my monitors are CC and I'm using a NVidia GeForce GTX 1060 6gb. Windows 10 Pro (but not the very latest update because I refuse to subscribe to that MS BS).

Just an observation: The masking would be much less visible in the nightclub if you use a more geometric path however that pesky green line is a complete deal breaker for any pro use.

BTW where is that nightclub?

[ungeherr]

Re: Livepano layer mask visible in chrome

by ungeherr » Thu Oct 26, 2017 7:35 am

Hi ERR,

Can you please take a screen shot of both the fan and the club so we can see if they differ from my screen shots.

I don't know if I could have drawn the path much differently without sacrificing resolution. In some cases it's a balance between the best path and resolution. In the livepano editor the video image looks like a time glass so I couldn't go much wider.

[benji33]

Re: Livepano layer mask visible in chrome

by benji33 » Thu Oct 26, 2017 10:51 am

@ungeherr With your samples, you're true. It's better on Chrome & Firefox.
I don't understand why I've not noticed that yesterday with the nightclub sample. Strange. Probably a file naming issue on my side. Sorry about that.

@ERR here we are not talking about the mask which in fact is the polygon defined but on the video encrustation and the "colorimetry".

[ungeherr]

Re: Livepano layer mask visible in chrome

by ungeherr » Thu Oct 26, 2017 1:00 pm

I would be lying if didn't say that this is a hallelujah moment. You finally believe me

[ERR]

Re: Livepano layer mask visible in chrome

by ERR » Thu Oct 26, 2017 3:28 pm

Screenshots from latest firefox

[benji33]

Re: Livepano layer mask visible in chrome

by benji33 » Thu Oct 26, 2017 3:38 pm

So I've read many different thing since I have answered this topic.

In fact some of the latest 64 bits browsers benefits now of the hardware accelerated video decoding. It's a good thing but ... there are using the YUV standard designed for TV rather than the RBG one designed for PC. And it's why I wasn't agree a few post before with you because there is no logic at all to do that.

On our side I don't remember if there was a change on the color range. I don't think so and i think the issue is now raised by the new version of browsers.

I've asked the developer, which is in charge to maintain our Video brick used into our products, a review of what is possible on the encoding into this brick with the specific encoding codec in it.
I can't tell about a quick correction or not but I've now all the needed to make the right report and tell about the needs into our Video encoding software brick.

Thanks again for your patience and for all the effort to help us to improve our software.

PS: the partial solution can be to convert video by your own in a first time for the high and low livepano versions generated.

[ERR]

Re: Livepano layer mask visible in chrome

by ERR » Thu Oct 26, 2017 5:18 pm

Seems like a plan. Thanks benji for your attention. Hopefully we can look forward to a "fix" soon.

[ungeherr]

Re: Livepano layer mask visible in chrome

by ungeherr » Thu Oct 26, 2017 8:11 pm

Thanks Benjamin! This really makes us super happy. Great support!

I agree that it makes no sense to use a limited color range. As you can see in the attached screen shots, the untouched livepano via chrome looks better if you disregard the visible mask.
But there must be a reason why safari renders the untouched livepano clip as limited.

What I find interesting with safari is that it manages to remove almost all masks completely. Safari does a better job than Chrome with livepano processed though hitfilm. This is probably out of your control to fix on your end but it should be interesting for you to know why Safari is so much better.

[signmaster]

Re: Livepano layer mask visible in chrome

by signmaster » Thu Oct 26, 2017 10:00 pm

Finding this all a bit hard to follow, could someone do a summary, especially on how to convert your video for “high and low livepano”

Steve

[ungeherr]

Re: Livepano layer mask visible in chrome

by ungeherr » Fri Oct 27, 2017 7:55 am

@signmaster
This is what you need to do until there's a fix from kolor
1. Download, install and start hitfilm 4 express
2. choose new, start editing (project settings comes later)
3. Locate the spots folder of your exported panotour project and drag/import the high livepano file to the media section.
4. Drag the livepano file to the timeline. You will be asked to change the Editors sequence settings. Press YES
5. Go to the "export tab" and then the "mp4 tab". Change the video and audio bitrate so it matches the original. Hit export.
For some reason hitfilm tends to use a higher bitrate than what you choose (at least for me it does) so use the following bitrate for the high and low livepano. You want to end up with the same or slightly lower file size so some times you have to lower the setting even more.
6. Go back to the edit tab and remove the high livepano and import the low livepano. Drag it to the timeline and you will be asked if you want to update the settings again. Press YES.
7. Go to export, mp4 and change the bitrate. Hit export

High livepano: Target bitrate - 3.9Mbps, Max bitrate - 4.5Mbps
Low livepano: Target bitrate - 1.9Mbps, Max bitrate - 2.1Mbps

[feyona]

Re: Livepano layer mask visible in chrome

by feyona » Mon Feb 19, 2018 2:16 pm

Hi, Are there any updates on this isue? I am having the same difficulties, and will be trying the Hitfilm express "fix" but have a few clients using livepano right now, and de looks + time it takes to fix them are far from optimal right now.

[ERR]

Re: Livepano layer mask visible in chrome

by ERR » Mon Feb 19, 2018 4:23 pm

I don't think Kolor has any plans at all to address this issue. They obviously don't think it's a problem otherwise it would have been fixed a long time ago.

[benji33]

Re: Livepano layer mask visible in chrome

by benji33 » Fri Feb 23, 2018 12:49 pm

It's a problem but our internal algorithm can't manage the color range like needed here at this time.
It's an opened issue on our side even if not solved yet. And I know, this issue is opened since a long time.