Colorblind and ChatGPT: I can finally follow pro pool

iusedtoberich · 2025-07-24T08:33:32-0700

It took 30 years, but I finally have a solution to knowing what the hell I'm watching.

It didn't get the labeled image correct, but I can probably bug it to fix it.

iusedtoberich · 2025-07-24T08:51:06-0700

This might be the tipping point for me to buy this.

freds · 2025-07-24T09:07:44-0700

That's pretty cool. Do either of these 2 on the right (also produced by chatgpt) somewhat match what you see for the left image?

iusedtoberich · 2025-07-24T09:29:39-0700

freds said:
That's pretty cool. Do either of these 2 on the right (also produced by chatgpt) somewhat match what you see for the left image?

View attachment 839467

The right image looks the same as the left to me. I can't identify all the balls on any of the 3 images (equally as bad on all 3).

freds · 2025-07-24T09:40:31-0700

Wow yeah that's tough. I guess you're used to checking the numbers. (Playing, not watching.)

Another reason to play straight pool and one pocket instead.

GorstyBoy · 2025-07-24T10:28:55-0700

Wow! This threw me for a fun image processing rabbit hole. For a stationary camera, you can mask off what pixels are in bounds for the table playing surface. Then use the hough transform to find all the balls based on how many pixels of radius a ball would fall within. Then tune high and low values for the RED GREEN and BLUE pixel values for each color ball. Combine all of that logic and you can identify most but the nine ball cleanly with no AI. The nine can be found with a little blob analysis of finding both the yellow and black in high enough concentration within a given pixel radius. The shadows all get to have a go at being the eight ball.