i have really anoying problem with my computer vision code that i just cant deal with it..
so same video and spin played over and over and different timings..
1236.38 ms
1242.84 ms
1235.83 ms
1248.97 ms
and for the early prediction im going for this is big error… So yeah if anyone had deal with similar problem in opencv and might have solution i will appreciate his help!
il edit and add timings for a faster ball:
736.55 ms
722.50 ms
737.94 ms
721.82 ms
Such difference has no influence on prediction results. Humans when they click do 50 ms+ reaction mistakes and still nothing bad.
Your program simply fixes mistakes which do your computer or device in which you run video - that you can’t avoid.
In what do you code?
Any idea using new technology:
Rather than relying on human rotor/ball clicks, and full revs of the ball and wheel, perhaps someday the ultimate hybird computer could be developed so that it could see the ball on the track and zoom in on it, much like the auto focus cameras do these days. The sensors could then automatically track the ball, and rotor, and adjust in real time perhaps. Such auto ball tracking would enable the computer to adjust to the ball shudder/shock waves by measuring fractional revs, several at a time.
Quote from snowman
Full revs can’t make you a big winner or even win. Fractional revs can allow you to predict the exact ball timing and maybe also the decay orbit pattern.
im using python, the core problem is that im trying to predict as early as first/second ball rotation
also my iteration times for the while loop are quite random i will try remaking it in c++ and hope its going to get withing acceptable range
Maybe you know if is possible to make a program with C++ which can work on Android?
Similar to programs done with Java…
c++ should be possible, i currently use java im working on android app that im going to share, first i need permission from forester xd
Dude you newd quarter rev, can your program do it? And correlate the relevant rev to the type of orbit decay pattern of the ball.
a program can do whatever you program it to do what is that question? i dont exactly understand what do you want it to do? you want to detect every quarter revolution of the ball? And than you want the decaying of the ball for every quarter rev?
Yes! Then correlate each type to different scatter because difference dealer and difference being method has different resulting scatter
I wrote my first program in C++ builder from Rad Studio. And the followup in Delphi from Rad Studio 12.3. The Delphi software is faster and more easier to program. If you don’t use a camera for the timing Delphi is ok.
that problem is long resolved xd testing stage is over so c++ is not in help i need javascript so it can be ran on multiple wheels and collect data automaticly.
How do you extract the roulette video live into a app/system? Asking for a friend…
What do you mean into app/system? 2nd grade with basic coding skill will figure it out. Real time screen capture or look how the page works and get the video stream
Its funny you replied now as i only just found out you can track the ball and its speed etc through real time screen capture..as for the box calculations-i will not be doing this, PINNs would to calculate the diameter and physics of the ball. Everything is mostly automated in terms if the logic/calculations. The labour is in the coding/implementation itself.
I concider my self a hobbyist and as someone that didnt lisen and also go through the ai pipelines phase, you just gonna lose your time.
All ai implementation is completed and running with 0 bugs but even more importantly, having an actual positive impact towards the final recommendation(s)
Delphi is a relic these days. People use C++, Python, and Java. Why? There are plenty of libraries for those languages, so if someone doesn’t already know Delphi, it’s pointless.
Automated timings will not help you that much, a good manual input with corrections has good accuracy. The automated thing is for online wheels if you want to play them. The casinos will switch cameras, so you can’t collect the data, but they still show enough so you can collect the data with a program, but not everything, some has to be done manually, it mostly depends how what they do to hide what’s going on.
I had a program in opencv(I say had, because it was something like pre-alpha, but it was working, i still have the code and some testing movies i did, but it will be rewritten / moved to a new system for image processing with ai this time), it was an image processor (NOT AI), a processor processes the images and calculates motion on the screen using math by generating contours and thresholds, but it’s not ai.
AI isn’t going to help much, if they switch cameras and the only way to know the drop point, is by the sounds the diamonds make(you can pick it up which diamond was hit, but you can’t see exactly, and half of the wheel is hidden), and if someone would like to make it for real casinos, it’s nearly useless since you still have to be there, so you might just as well do it manually, which is more reliaiable and more “trustwordy”.
I am mostly focused on the actual prediction program, data collection system, data analysis, timing corrections, diamond hit predictions, ball scatter, rotor input methods, and other data input methods, to make them as simple as possible while maintaining what was going on originally. Some variables you can predict by other ones, so a lot of stuff can be reduced, so it’s applicable.
I am wondering what you two specifically got already, because you talk about auto-data collection from the wheel. I assume it’s for late bet wheels, with computers being applicable there (it might as well be some other method that you bet before the spin, but these are harder and have a lower edge), so what do you have? Do you predict diamond hits? Is the scatter collected for each diamond or merged together? Are the rotor speeds isolated? Do you have ANY corrections for manual timings?