All computer vision stuff is like
Step 1. Do some linear algebra that makes logical sense, uses a rational scientific model of how we believe the world works, looks reasonable
Step 2. The result of Step 1 is basically garbage. Feed it into a nonlinear optimizer to make it actually good
About
"The program should, in some way, expand the concept of what computer paint programs are, as well as what mark making can be."
stubbornly calling it `rotationVectorToRotationMatrix` instead of `Rodrigues`
Also wondering if this technique would help with RFID stuff -- I hadn't realized that there is this same problem of discrete ambiguity even in visual marker tracking https://social.omar.website/@omar/statuses/01J0F5H0GW4HMG3G6B5RR0TQ9T
Been reading about "Virtual Visual Servoing" http://rainbow-doc.irisa.fr/pdf/2002_eurographics_marchand.pdf -- maybe the most convoluted way to explain a smoothing/filtering algorithm (?) that I've seen lol
What is a car mouse
(I'm sure people who've actually used CAD software know this:)
my current understanding is that both clang and gcc basically support nested functions / closures but with different and incompatible syntax and implementation… wonder if people ever just ifdef between them
Have ppl ever used `git notes`...