/tech/ - Technology★

>>885369 (OP)

>finding people to try this problem

Unironically, stackoverflow.

>>885379

but stackoverflow is strict on the "questions and answers only" format, ill probably get downbumped if i start a challenge thread

seems like the box is an unnecessary complication, the images could be pre-generated, then your looking at general OCR. You don't need people to try to solve this problem, you just need various algorithms, free and proprietary software to try to solve the problem, then compare the performance.

>>885402

seems like the die is an unnecessary complication, the rolls could be pre-generated, then your looking at general OCD. You don't need to try to solve this problem.

>>885402

>>885405

well the whole point of the test really would be to extrapolate and identify text from a real life image

we could also have it so the camera is not neccessarily looking straight down at the die, just from an angle where a human would also be able to identify the number rolled

the main idea here is that the difficulty of the problems stems from being able to make out which number is on top (rolled) and being able to extrapolate and interpret text data

so only half of it is really ocr

>>885398

There's also codegolf.stackexchange.com, perhaps that's a good place to post it?

>>885473

still doesnt look like the right place to post it, but this looks pretty cool for casual code challenges

thanks anon

>>885828

You might also want to try mathematica.stackexchange, and https://dsp.stackexchange.com

This example below is quite cool, it's from the regular stackoverflow, but you'll find a lot of people willing to complete similar change type questions on the above too. In 2011, SO wasn't as crap as it is today.

https://stackoverflow.com/questions/8479058/how-do-i-find-waldo-with-mathematica

>>885835

*above two.

>>885369 (OP)

>this die would be in a transparent box with a camera above it looking down

>the box would be shaken, rolling the die

>the camera would then take a picture

>the goal for any algorithm attempting this problem would be to determine the number that was rolled

This was already done as a student project a decade ago.

>>885838

are you shitting me

There's scikit for python, or if you want to do it by hand you can too, although with a 20-sided die it'll be nowhere as easy as with one with the dots. Anyway, choose some image recognition lib and try it, make sure the dice can be seen clearly by the camera.

>>885838

Correct.

>>885842

i was thinking openCV

>>885842

the serious problem here is to isolate the text from the top, because 9 other sides are visible at any given time

>>885852

But you know the text beforehand. The one on top should be the best match since the others are at an angle. You could also easily cut the top since you know the size of the die.

>>885853

what about from an angle or a distance

>>885855

The distance is also a non-problem since it's known (even if the die isn't dead-center in the box the distance to the camera is pretty much the same). As for angles I don't know, I only used ARToolkit I think it was and it uses/used some patterns so they were easy to identify, worst case you could do the retard solution of rotating it and picking the best.

>>885838

>This was already done as a student project a decade ago.

Was it the one with a bunch of dice being rolled at once?

>>885894

Character recognition. It has been done many times before.

What is this, 1970? This is a trivial problem.

>identifying the die in field: trivial

>correcting for angle: trivial

>isolating top face: trivial

>rotating text to be horizontal: trivial

>doing ocr on the text (and repeat after rotating 180 degrees): trivial

In fact you don't even need to do ocr, since there's only 20 possible sides. Just take a photo of each and match to them. Or even more brute force way: For each side, take 100 photos of die from different angles, for a total of 2000 photos, then just find the most similar one among those with minimal processing. And what's this dice camera faggotry? Just generate it with a script. It's just a shitty captcha anyhow.

>>885982

Probably PRNG. Still, OP is a fag

Paint sides in different colors, detect color in horizontal plane. Are you trying to reinvent low-entropy PRNG?

>>885995

>>885894

>>885982

>>885987

>>885995

well im getting the sense that this has already been though of and solved

lets shift gears and just talk about image recognition in general

>>886253

thought of*

>>885839

https://www.youtube.com/watch?v=95EErdouO2w

>>886366

thats actually mighty impressive