So if I'm understanding correctly it's doing a spherical projection onto a plane, we get to navigate that plane, and every snippet of that plane-projection we see is being projected back into 3D somehow?
It can't actually reconstruct the 3D environment from the projection, so what's actually happening there at the end?
So if we navigate the plane with the projection of the sphere on it and apply the reverse transform to the snippet we're viewing (genuine question: what form and size would the snippet need to have?) it seems we would get the effect of being inside the sphere.
I don't know the form and size of the snippet, I'm assuming it's rectangular because I don't see a reason for it not to be..
I think We're still having an issue at the poles: If you take a rectangular snippet out of the image you've uploaded which its topmost edge coincides with the topmost part of the image, what can we expect the transformation to yield?
They might be dealing with edge cases behind the scenes in some way. Mathematically speaking there can't be a way to fully invert a transformation from 3D to 2D, so I guess they settled for something that they can still work around.
In any case even though I'd like to know the exact details, at this point my curiosity seems to be satiated.
5 comments
[ 3.9 ms ] story [ 21.8 ms ] threadI have a couple of ideas but I prefer informed answers over my guesswork.
Example screenshot: https://i.imgur.com/z2fwIzJ.png
So if I'm understanding correctly it's doing a spherical projection onto a plane, we get to navigate that plane, and every snippet of that plane-projection we see is being projected back into 3D somehow?
It can't actually reconstruct the 3D environment from the projection, so what's actually happening there at the end?
So if we navigate the plane with the projection of the sphere on it and apply the reverse transform to the snippet we're viewing (genuine question: what form and size would the snippet need to have?) it seems we would get the effect of being inside the sphere.
I think We're still having an issue at the poles: If you take a rectangular snippet out of the image you've uploaded which its topmost edge coincides with the topmost part of the image, what can we expect the transformation to yield?
They might be dealing with edge cases behind the scenes in some way. Mathematically speaking there can't be a way to fully invert a transformation from 3D to 2D, so I guess they settled for something that they can still work around.
In any case even though I'd like to know the exact details, at this point my curiosity seems to be satiated.
Thanks for the friendly conversation. :)