@andrewPolygon@FreyaHolmer@GeminiApp Yes, so basically that's what the visualization shows. When a line becomes much thinner than 1 pixel, coverage becomes unstable because many pixels no longer sample enough of the triangle/line area consistently, so the rasterized result starts to jitter or disappear.
I was trying to understand how @FreyaHolmer Shapes library draws line geometry so crisp at any distance and had @GeminiApp explain it to me. I was not expecting this visualization
@JomaaMiro@FreyaHolmer@GeminiApp Yes, itβs a form of AA implemented directly in the fragment shader. Traditional AA techniques are often done either through multi/super sampling during rasterization or as a post-processing pass
@toncijukic@FreyaHolmer@GeminiApp The exact one I shared is in a longer conv I can't share but really you just need to ask for an explanation on the topic https://t.co/Dfly8SNMDP
@lisyarus As a geology nerd, I have to point out that rivers usually donβt meet at very sharp angles, they tend to join at roughly 90 degrees. Even if two tributaries start off meeting at a tighter angle, sediment buildup will gradually widen it over time.
@marcsh The reality is more nuanced and even more fascinating. Rail engineers design train tracks using a centerline clothoid but here's the catch: the offset of a clothoid is not a clothoid itself. So left and right tracks do not trace a perfect mathematical clothoids
@marcsh Yeah, NURBS are an interesting topic on their own. It's fascinating how by nature they can actually accurately represent circle arcs while bezier splines can only aproximate them.
@ergocortex Yes, real engineers typically use clothoids as the main road shape for high speed roads like highways or national roads, but even there lanes don't run truley parallel and they are just aproximations in practice. For residential and city infrastructure they rely on arcs.