Skip to main content
Participant
July 5, 2016

Cube Vs Sphere for Pano projection

I have always used a sphere with inverted normals to project my equirectangular panos and show them in a VR scene with the camera in the center of the sphere.


Another option is to use a cube with a Skybox/cubemap material to project the same pano (converted from equirectangular to cube facing).


Visually I see the same quality, but in terms of performance, with the cube I only have 12 tris, instead of the large number of triangles that I get with a sphere.


My question is… ¿should I always use cubes and forget to use spheres to show panos?


Thanks for your comments

This topic has been closed for replies.

4 replies

Inspiring
July 5, 2016
should I always use cubes and forget to use spheres to show panos?
Not a bad idea. Equirect has some nice properties, such as the way it looks on a flat rectangle, but cubmaps are easier to render, they distribute the pixels better, and they have no singularities. EleVR has a nice post on the subject.
July 5, 2016
Definitely use cube: faster rendering, better storage and suitability for compression e.t.c. . We use cube maping in DepthQ VR renderer for 380 degere 3d vido with succes. But you must be careful and precise on edges ....
Up and Coming
July 5, 2016
You are in a mathematical object, you can render a full screen quad or use compute shaders to raycast on it, no need for a real geometry in the first place. But if you are worried about the triangle count, be not, GPUs are made to hide latency in their pipeline, even with a few thousands triangles, the cost of filling the screen with a texture is quite enough to turn the vertex cost to absolutely nothing.
p4bloAuthor
Participant
July 7, 2016
Thanks for your answers, the explanations were very clear.


galopin said:

you can render a full screen quad or use compute shaders to raycast on it, no need for a real geometry in the first place. 


Galopin, I'm curious to know more about it. Could you recommend me some reading about it?