Lecture 13: Global Illumination & Path Tracing (31)
sberkun
Fun fact: this model is called the Stanford bunny, and is a famous test model. You may also recognize the Utah teapot from assignment 2.
orenazad
https://www.youtube.com/watch?v=TIxt9guMbXo
Tom Scott has a great video about the Utah teapot and it's relevance throughout the history of Computer Graphics!
jacklishufan
In previous lectures we talked about the light decays as it transmits through space and different lights may have non-uniform decay that depends on directions. Why are we not considering the distance between p and p' in this equation
ChangyiYang
I don't quite understand why the radiance invariance works, and the meaning behind it. Is that saying there is only one radiance value in one ray direction?
sberkun
As context for this slide: If we wanted to simulate light accurately, we would simulate all the rays emitting from all the light sources, simulate all their bounces, then add up the ones that hit our "camera". However, this ends up wasting 99.9999% of our work, since only a tiny fraction of light ends up at our camera.
To be much more efficient, we trace rays "backwards", tracing ray paths out from the camera until they hit a light source. Radiance invariance says that tracing the light "backwards" from p to p′ will give us the same result as tracing the light "forwards" from p′ to p. Therefore, the much more efficient "backwards" method doesn't sacrifice any realism.
LeslieTrue
Curious about the specific form of the transport function. Usually how do we design it?
Fun fact: this model is called the Stanford bunny, and is a famous test model. You may also recognize the Utah teapot from assignment 2.
https://www.youtube.com/watch?v=TIxt9guMbXo
Tom Scott has a great video about the Utah teapot and it's relevance throughout the history of Computer Graphics!
In previous lectures we talked about the light decays as it transmits through space and different lights may have non-uniform decay that depends on directions. Why are we not considering the distance between p and p' in this equation
I don't quite understand why the radiance invariance works, and the meaning behind it. Is that saying there is only one radiance value in one ray direction?
As context for this slide: If we wanted to simulate light accurately, we would simulate all the rays emitting from all the light sources, simulate all their bounces, then add up the ones that hit our "camera". However, this ends up wasting 99.9999% of our work, since only a tiny fraction of light ends up at our camera.
To be much more efficient, we trace rays "backwards", tracing ray paths out from the camera until they hit a light source. Radiance invariance says that tracing the light "backwards" from p to p′ will give us the same result as tracing the light "forwards" from p′ to p. Therefore, the much more efficient "backwards" method doesn't sacrifice any realism.
Curious about the specific form of the transport function. Usually how do we design it?