comments 4
NERD WARNING: MY SUPER POWER IS OVER THINKING THINGS. SCROLL ON BY IF THIS ANNOYS YOU. I'm familiar enough with AI to know the process of how it teaches itself based on training data and trail + error until it develops a consistency and accuracy but that it cannot explain to the user how it came to the correct conclusion but... I can't help but wonder about how it decides which edges are correct and which are not. So, given three visible edges of a cube that are parallel in perspective and therefore should converge to the same vanishing point, there are 3 situations. 1 All do properly converge to the same vanish point(easy for AI to check) 2. Two incline toward one another but the third does not(also easy for AI to identify which is the incorrect edge) . 3. They all incline toward each other but don't all converge to the same point. The third case is the most interesting. If all three edges incline toward one another but do not converge to the same point, then which two edges are correct and which one is not? Any two of the will intersect if extended far enough. So which two define the correct vanishing point? I'm sure there's not an easy answer and that Stan can't share the details of what's going on under the hood but it's fun to think about these things.
LESSON NOTES
How do you study perspective? Marshall breaks down the elements of perspective and how you’d go about learning it. Stan also challenges Marshall to draw some boxes and run them through his new Box AI BETA App that’s being developed in the Proko lab.
“Draftsmen” is available in audio. Subscribe on these platforms to keep up to date:
Spotify, Stitcher, Apple, Google
Referenced Artists/Works:
Perspective Drawing Handbook by Joseph D'Amelio
Perspective Made Easy by Ernest Ralph Norling
How to Use Creative Perspective by Ernest William Watson
Kim Jung G - Omphalos
Perspective! for Comic Book Artists by David Chelsea
Framed Perspective Vol 1 & 2 by Marcos Mateu-Mestre
Glenn Vilppu
Marshall’s old perspective course
COMMENTS
NERD WARNING: MY SUPER POWER IS OVER THINKING THINGS. SCROLL ON BY IF THIS ANNOYS YOU. I'm familiar enough with AI to know the process of how it teaches itself based on training data and trail + error until it develops a consistency and accuracy but that it cannot explain to the user how it came to the correct conclusion but... I can't help but wonder about how it decides which edges are correct and which are not. So, given three visible edges of a cube that are parallel in perspective and therefore should converge to the same vanishing point, there are 3 situations. 1 All do properly converge to the same vanish point(easy for AI to check) 2. Two incline toward one another but the third does not(also easy for AI to identify which is the incorrect edge) . 3. They all incline toward each other but don't all converge to the same point. The third case is the most interesting. If all three edges incline toward one another but do not converge to the same point, then which two edges are correct and which one is not? Any two of the will intersect if extended far enough. So which two define the correct vanishing point? I'm sure there's not an easy answer and that Stan can't share the details of what's going on under the hood but it's fun to think about these things.







