Art : AI vs Human III Colors

Hello everyone, I hope you're all doing fine. This post is a continuation of "AI vs. human art" you can check the previous post here.

But before we dive in, I will indeed bore you with a big paragraph about painting because I can.

So let's talk about colors. What are they exactly? If we're taking the RGB model, which obviously stands for a really good bong, and not red, green, and blue, which are the primary colors as we all know in elementary school and every color can be made by mixing these 3 (the problem with this model is we can't compare them easily, and that's where the HSB model comes in),

and since this model uses a multitude of different colors, it is harder for humans to use it, as it is more suited for technical devices like your monitor or phone screen. But there is another model, which is HSB, and it stands for hue saturation and brightness.

To explain what hue is because it's a weird word, it is the shade of a color, or in a more nerdy way, the wavelength within the visible-light spectrum at which the energy output from a source is greatest, aka pure color without a tint.

Then we have saturation, which is the intensity of the hue, and brightness, which is how much black or white is in the hue, and as you can tell already, it's way easier to look at a color and tell approximately what it's composed of in this model differently from the RGB model, which is confusing for human eyes, so if we're using AI, why should we pick a model that's better for humans? We can just use RGB and the AI will manage, I hear you say.

Well, that's fair enough, but we want the AI model to use our human standards when detecting these flawless shapes. Another reason is due to the way we're going to detect these shapes, which is edge detection it is a method of segmenting an image into regions of discontinuity.

Edge detection is a widely used technique in digital image processing. And this takes us to why I was talking about different color modeling. As I mentioned in the HSB model, b stands for brightness, and every color has unique digits for his HSB identity, so naturally, if let's say two colors make contact with each other (colors are a spectrum of shade, so blue and a slightly more intense blue are two different colors), Edge detection will, you guessed it, detect that discontinuity and highlight the shape, as seen in the example down below* (I couldn't find the name of the artist for credits; I will be using my paintings for future posts). If you're more interested in how colors interact with each other and how different colors create lines when they make contact with each other, you can check out Color Theory.

So we are able to detect shapes in a painting. Okay, now that didn't solve the problem, you're saying, and you're right, that's where years of sketching and pages of my sketchbooks will come in handy, but I will talk about that in the next post.

*The Prewitt filter was used to demonstrate the results of edge detection.