Computers use rectilinear perspective projection (divide by camera-space z). The one that more closely models the scenario Jon gives is curvilinear (divide by distance from viewer). The former has the essential property that lines project to lines. So your quads, triangles, etc remain polygons so we can still use the triangle rasterizer. Cameras produce the kind of projection Jon talks about because they have rectilinear lenses that intentionally distort the image to reduce curvilinear effects. Your brain does voodoo magic including inverting the image you’re seeing and distorting it so doesn’t look like GoPro footage. I’m not an expert in this stuff, but I am kind of bothered Jon glossed over it so I had to say something.
When you say curvilinear does that refer to the curvature of our eye and how it effects the image our brain produces as a result? Really curious as i just started learning 3D math for graphics and from i can tell, i think you are correct
@@Leeszus My bad. Curvilinear isn’t that technical of a term. Basically what I was trying to say is if you use a lens system like our eyes in a camera and digitally image what hit’s the sensor you’d get something that looks like gopro footage. To avoid this your brain does a lot of processing so that you experience something like a rectilinear perspective projection. Camera’s try to produce images like our brain does (rectilinear perspective projection) so it looks natural. They just use special lenses to do this instead of signal processing like the brain does. Our retina’s being spherical might reduce the effects of the distortion, but I think it’s still there. These are just things you have to think about. If you’re just concerned about getting something working the math is surprisingly easy for rectilinear perspective projections. OpenGL and graphics API’s in general muddy the water though.
Wait, so some of the light that comes off my monitor and hits my eye actually after it left my monitor traveled around Jupiter, bounced off a mirror in china, and then sketched out my name in the sky before finally making it to my eye?
I wish all the efforts in AI development would make the camara autofocus work one day instead of boring chat-bots... No one ever solves that problem correctly, it angers me.
Computers use rectilinear perspective projection (divide by camera-space z). The one that more closely models the scenario Jon gives is curvilinear (divide by distance from viewer). The former has the essential property that lines project to lines. So your quads, triangles, etc remain polygons so we can still use the triangle rasterizer. Cameras produce the kind of projection Jon talks about because they have rectilinear lenses that intentionally distort the image to reduce curvilinear effects. Your brain does voodoo magic including inverting the image you’re seeing and distorting it so doesn’t look like GoPro footage. I’m not an expert in this stuff, but I am kind of bothered Jon glossed over it so I had to say something.
When you say curvilinear does that refer to the curvature of our eye and how it effects the image our brain produces as a result? Really curious as i just started learning 3D math for graphics and from i can tell, i think you are correct
@@Leeszus My bad. Curvilinear isn’t that technical of a term. Basically what I was trying to say is if you use a lens system like our eyes in a camera and digitally image what hit’s the sensor you’d get something that looks like gopro footage. To avoid this your brain does a lot of processing so that you experience something like a rectilinear perspective projection. Camera’s try to produce images like our brain does (rectilinear perspective projection) so it looks natural. They just use special lenses to do this instead of signal processing like the brain does.
Our retina’s being spherical might reduce the effects of the distortion, but I think it’s still there. These are just things you have to think about.
If you’re just concerned about getting something working the math is surprisingly easy for rectilinear perspective projections. OpenGL and graphics API’s in general muddy the water though.
interesting
On the video preview Jonathan himself rendered very well
For the rest of the video, search:
How 3D video games do graphics. Jonathan Blow
czcams.com/video/bGe-d09Nc_M/video.htmlsi=IAtTTMw2jFwAi6eP&t=1251
"4$k computer and I can't do anything with it" so succinct
Wait, so some of the light that comes off my monitor and hits my eye actually after it left my monitor traveled around Jupiter, bounced off a mirror in china, and then sketched out my name in the sky before finally making it to my eye?
I wish all the efforts in AI development would make the camara autofocus work one day instead of boring chat-bots... No one ever solves that problem correctly, it angers me.
Sony?