In case anyone is wondering, the two triangles are not the same color
Right is #D90000, left is #C11F03, no visual illusion going on here
muuncakez
so to prevent this change in color we would need to pick a smaller blur rate such that it's not impacted by the white bg... is the present color difference (and blur rate) also partly due to the screen resolution as someone in the audience had asked?
Staffi-geng
The two triangles are supposed to be the same color, except at the edges of the blurred triangle! If we were performing supersampling, for a pixel that is fully within the triangle, the average of the NxN samples will be equal to the original color of the triangle. For a pixel that is half inside the triangle and half outside the triangle, we would average our NxN samples are get a light-reddish color at the edge.
123453ANSH
What I found particularly interesting about this slide/concept presented on this slide is just the idea of rasterizing a triangle; being able to break a triangle/shape down into a point wise representation to me is just very interesting, and I'm sure will have more relevance later in the course as we start to work with different/multiple objects.
Hsong159
This slide reminded me of convolutional neural network. Just as filters in convolutional neural network, the color of the dots looks to be the average of their surroundings. The more white dots it is next to, the more shallow it looks. The red dots in the middle stays red because its neighbors are all red.
Staffyirenng
@buggy123 -- thanks for pointing this out. A student noted this in lecture. As @i-geng noted, the triangles are meant to be the same color away from the blurry edge. I've made a quick fix in the slides so that these are the same RGB colors, and they are updated here (thanks TA Connor!) now, to try to minimize confusion about that.
Hmm... interestingly, after going from slide export through ImageMagick processing on our server to produce the slide image on the left here, I noticed that a very small RGB difference crept in again between left and center triangles. Please ignore!
kbrgl
This reminds me of how text is rendered on the web.
There's a CSS property called -webkit-font-smoothing (on WebKit based browsers) with three options:
antialiased
subpixel-antialiased
none
If you’re curious how these look, check out this MDN page:
In case anyone is wondering, the two triangles are not the same color
Right is #D90000, left is #C11F03, no visual illusion going on here
so to prevent this change in color we would need to pick a smaller blur rate such that it's not impacted by the white bg... is the present color difference (and blur rate) also partly due to the screen resolution as someone in the audience had asked?
The two triangles are supposed to be the same color, except at the edges of the blurred triangle! If we were performing supersampling, for a pixel that is fully within the triangle, the average of the NxN samples will be equal to the original color of the triangle. For a pixel that is half inside the triangle and half outside the triangle, we would average our NxN samples are get a light-reddish color at the edge.
What I found particularly interesting about this slide/concept presented on this slide is just the idea of rasterizing a triangle; being able to break a triangle/shape down into a point wise representation to me is just very interesting, and I'm sure will have more relevance later in the course as we start to work with different/multiple objects.
This slide reminded me of convolutional neural network. Just as filters in convolutional neural network, the color of the dots looks to be the average of their surroundings. The more white dots it is next to, the more shallow it looks. The red dots in the middle stays red because its neighbors are all red.
@buggy123 -- thanks for pointing this out. A student noted this in lecture. As @i-geng noted, the triangles are meant to be the same color away from the blurry edge. I've made a quick fix in the slides so that these are the same RGB colors, and they are updated here (thanks TA Connor!) now, to try to minimize confusion about that.
Hmm... interestingly, after going from slide export through ImageMagick processing on our server to produce the slide image on the left here, I noticed that a very small RGB difference crept in again between left and center triangles. Please ignore!
This reminds me of how text is rendered on the web.
There's a CSS property called -webkit-font-smoothing (on WebKit based browsers) with three options:
If you’re curious how these look, check out this MDN page:
https://developer.mozilla.org/en-US/docs/Web/CSS/font-smooth