Post process issue with VR Camera.

[tranlong021988]

hi everyone, it's seem that an issue exist with VR stereoscope in WebVRFreeCamera and VRDeviceOrientationFreeCamera. 

Please check here: http://www.babylonjs-playground.com/#KBWBU

With this example, I just apply a post process to WebVRFreeCamera to modify brighness and contrast, but focus line 22, you can comment that operation to see what happen when apply post process to VR stereoscope. It seem that it give me wrong render result. 

Sorry for my poor English.

[JCPalmer]

I just switched to a FreeCamera for your PG.  The brightness are the same as with your example.  Do not understand the line 22 comment.  If you do not attach a post process, nothing you just get things un-modified.

[chicagobob123]

it looks like rendering order issue. You need to do this preprocess, 

if there is such a thing. Modifying the brightness post has impacted  the alpha channel which controls the antialiasing, 

which ruins the edge definition. Maybe it is a bug? I don't think the alpha should be effected by brightness adjustment. 

 

 

[Sebavan]

Hello,

The thing is until webgl 2.0 Texture Render Target does not support Anti Aliasing and you chosed to render in 0.25 so 1/4  (0.25 on line 9) of the screen creating the blocky effect.

In BJS post process are applied by first rendering to texture render target.

So by changing the render target ratio from 0.25 to 1 it removes the blocky effect but does not get rid of the aliasing effect:

http://www.babylonjs-playground.com/#KBWBU#4

This does not get rid of it fully due to Webgl limitations. To reduce it more you could also apply an Anti Aliasing Post Process of your choice.

There is nothing to do with transparency as those meshes are fully opaque.

[tranlong021988]

@JCPalmer: Let say I want to make a game, here is it(mobile browser only): https://dl.dropboxusercontent.com/u/86585940/BabylonWave/optimized5/index.html

When JetSki hit the 7Up can, I wanna make a bright screen effect applying a post process to camera. But it's seem that VR viewports between "before" and "after" not consistent. That issue calculate wrong projection distort. Can you see different result when commenting in/out line 22 ? 

I have record issue in action:

@chicagobob123 @Sebavan : It's about wrong projection distort, not about alisas. 

[Sebavan]

Sorry about the confusion,

It seems to be an issue with the reusable post process. I ll look at it later.

During this time, as it does not seem you need to reuse it, please change the reusable parameter of the constructor to false.

http://www.babylonjs-playground.com/#KBWBU#6

 

[tranlong021988]

@Sebavan : Sorry I don't understand much about post process but if I just let that parameter false, nothing affect my game performance ? 

[Sebavan]

Will be better cause only one render target :-) and you do not need the second one if you are not reusing it.

[MasterSplinter]

I was able to get this working correctly today.  We used webvr poly fill with a strict implementation of the api and had to create our own device database because the Google one they have is incorrect.  Unfortunately we have no method of obtaining a phones dpi using javascript so it needs to be done through user agents.  But, most of the data on devices is incorrect.  Side note gear vr  takes advantage of the api unfortunately their device orientation prediction is terrible and introduces all sorts of latency.

 

Trying to get in touch with a Boston Samsung team that's working on this...