rune solver not working for some runes about auto-maple HOT 7 CLOSED

Yea unfortunately they're bringing back some of the old colors. The algorithm currently filters the arrow colors first then does classification, so any colors that are complementary to yellow/orange won't show up. I'll try to find a more general approach than this, but yes more samples would always be greatly appreciated! Thanks for offering to help!

from auto-maple.

do you prefer a specific format for the images, besides the same aspect ratio as the bot? (low background quality, etc. )? I was also planning on taking them with the maplestory in game screenshot functionality.

from auto-maple.

I'm also having some issues but the capture.py isn't finding runes at all for me. I've added some logging and I can see that it's trying to detect but even with testing threshold down as far as 50%, it isn't ever finding the rune. I've tried both in 1920x1080 and the default 1366x768 that you had used and had no luck with either. Any advice on the detection never finding a match?

from auto-maple.

I'm also having some issues but the capture.py isn't finding runes at all for me. I've added some logging and I can see that it's trying to detect but even with testing threshold down as far as 50%, it isn't ever finding the rune. I've tried both in 1920x1080 and the default 1366x768 that you had used and had no luck with either. Any advice on the detection never finding a match?

This is a different issue, so I've made a new issue for it. #25

from auto-maple.

I don't really have a preferred format as long as the arrows are visible. But samples where the arrows kind of blend in with the background (hard for even humans) would be super useful in stress-testing new models. Of course, this is a lot rarer so please don't actively search for them, but when they do turn up, a quick screenshot would be really helpful!

from auto-maple.

I altered the bot to take a screenshot everytime it tries to solve a rune. So I still dont have many samples yet, and only like 1/5 of the samples i have so far are of the blue / green type.

How many samples do you think is necessary?
Also idk what type of algorithm youre using, but do you need me to label the samples?

from auto-maple.

No worries! I've been collecting some too, I think maybe 20-ish more samples would be enough to start prototyping with, and there's no need to label the samples. Considering the weird colored arrows are rarer, augmenting the existing arrow data by inverting/tuning the colors might actually be better than screenshotting by hand though.

from auto-maple.