A BYOAC Lightgun, are we up to it?

[rdagger]

Since, my cab has an LCD monitor I am going to go with a camera approach.  The field of computer vision is developing very quickly.  It should be possible for a computer with a camera to discern 2 guns and determine where they are pointing.    Instead of using LED tracking, the software could be trained to learn a gun's shape.  This could be a very inexpensive solution, because it is all software based.  You could use any webcam and you could use any gun.  You could even use a real gun (just take out the bullets and wire the trigger to an encoder or use blanks for realistic recoil)

Judging by the posts it seems like someone needs to start a BYOAC Lightgun website.

[Trimoor]

It can be done entirely in software, but that means more computing power is diverted from MAME.  Tracking 3 points of light takes far less power than object recognition and orientation.  The LED triad can probably fit inside the barrel, using a lens to spread them out.  If the tracking software is good enough, you wont even need to pulse them.  For the second gun, use four LEDs.

[sWampy]

Since, my cab has an LCD monitor I am going to go with a camera approach.  The field of computer vision is developing very quickly.  It should be possible for a computer with a camera to discern 2 guns and determine where they are pointing.    Instead of using LED tracking, the software could be trained to learn a gun's shape.  This could be a very inexpensive solution, because it is all software based.  You could use any webcam and you could use any gun.  You could even use a real gun (just take out the bullets and wire the trigger to an encoder or use blanks for realistic recoil)

Judging by the posts it seems like someone needs to start a BYOAC Lightgun website.

http://sourceforge.net/projects/opencvlibrary/  Has most of the code already written to do this, just slap an A or whatever on the end of one gun and a B or whatever on the end of the other, and you could be pretty set.   You could even put led's in the gun where they were wireless if you wanted. 

[sWampy]

It can be done entirely in software, but that means more computing power is diverted from MAME.  Tracking 3 points of light takes far less power than object recognition and orientation.  The LED triad can probably fit inside the barrel, using a lens to spread them out.  If the tracking software is good enough, you wont even need to pulse them.  For the second gun, use four LEDs.

You don't need the LED's in the gun.   Worst case if you can't track the gun easily, put 3 reflective dots on one gun, and dots in a different pattern on the other gun, and point an ir led back at the player to get some light on them. 

[ErikRuud]

Here is another head tracking application. http://freelook.org

They specifically state that you can use an LED as a marker.

[1UP]

You could use any webcam and you could use any gun.

[ErikRuud]

Yes, three of four points is more than enough.

[Trimoor]

So, anyone started on a test program/drivers?

[Flinkly]

not to trample on the effort, but if andy is doing well on creating his own for the comunity, wouldn't it be kinda a waste to make our own? he would probably be done before we got too far anyways, just hate to see you guys waste your time.

p.s.  i'm not trying to trample on your guy's feet, so if you don't see it as a waste, go on and make one to prov me wrong. 

[Adam]

I had a shot at a reverse webcam lightgun tracking system.
I basicly threw together the webcam and squares detection components of opencv.

http://users.tpg.com.au/mindle/camdemo.rar

Point your webcam at the screen and hit a key to take a pic. Escape quits.
You need the whole screen to be visible for it to work. If grabbing the picture with a decient webcam is fast enough i'll probably move to using a few LEDs around the screen.

[Silver]

not to trample on the effort, but if andy is doing well on creating his own for the comunity, wouldn't it be kinda a waste to make our own? he would probably be done before we got too far anyways, just hate to see you guys waste your time.

p.s.  i'm not trying to trample on your guy's feet, so if you don't see it as a waste, go on and make one to prov me wrong. 

I agree... I think some time and effort would be better spent improving the mame code for light guns or perhaps a driver that supports raw input and not mouse emulation. I  think this is how dual lightguns are forced to work already.

Such improvments would benefit both DIY webcam projects - the TrackIR system already does not use mouse emulation - an other setups.

[NiN^_^NiN]

http://www.arcadecontrols.com/wwwboard/messages/40217.html

I'm tired so i didnt look if it was posted so kill me if it was 

[Lilwolf]

my webcam is crap.

but isn't it true that they just aren't fast enought (forget about the processing time per image) to be used?

it seems like you might be able to get it to work on the first level or two on some shooting games if you got the drivers working... but after that... the cameras wouldn't take pictures fast enough for the computer to figure out where the pointer SHOULD be looking

and after that... even if everything above worked....

how would you make a trigger press?  thumb twitch?  or middle finger?  seems like this would be another huge hurtle.

I personnally think it would be super cool.  I would love to see someone use this to make a DDR pad (on any surface)... and a few other items like that... but I just don't see it being a real solution for gaming...  just something to show your friends...  "no really... just point at the screen and say bang... it really works!"

[Trimoor]

Most webcams can take images at a full 30fps.  Some might be able to achieve greater speeds.  With enough processing power, it can track the guns in real time.

Games run at 60 fields per second interlaced, and I'm guessing that most of them only render 30 frames per second, one image for both fields.  I think it will be plenty fast.

As for the triggers, they would be connected to a keyboard encoder.  If wireless shooting is desired, the triggers could activate an IR LED in the front of the gun for the camera to see.

[Silver]

I think most webcams would not be up to the task. Depends how good you want it....

The TrackIR system above is designed for use in games (head-panning in flight simulators etc...) and the premium edition runs at 120fps - the slowest one is 60 or 80fps I think.

They get around the processing image overhead by using infra-red emitters to reflect a very obvious point source, making it very easy to follow. they state cpu usuage for realtime tracking is under 1%. I am sure this would work very well as a gun - but again its a custom driver or mouse emulation. And if you want two guns you would probably have to program it yourself. Although it looks like they have a SDK for their optitrack or whatever....

[ErikRuud]

not to trample on the effort, but if andy is doing well on creating his own for the comunity, wouldn't it be kinda a waste to make our own? he would probably be done before we got too far anyways, just hate to see you guys waste your time.

p.s.

[Lilwolf]

Well, if processing power is then the issue... heres a question....

Could we take that software and add it to a seperate computer.. Then pipe the output out and into another computer through another machanism?

It seems like if we could find a dedicated computer solution... then it wouldn't effect the original computer... It might make this doable...  Plus you wouldn't have to worry about the game being emulated to have it work well.

I would try to find a 1+ghz machine for a dedicated solution if it worked well.  Especially if it coudl be headless.

And after writing the last message... I've now decided it needs to have speach recognition software installed... so that saying bang would be enough.... I really think telling someone "just point and say bang" would be by far the coolest thing!

[ErikRuud]

I had a shot at a reverse webcam lightgun tracking system.
I basicly threw together the webcam and squares detection components of opencv.

http://users.tpg.com.au/mindle/camdemo.rar

Point your webcam at the screen and hit a key to take a pic. Escape quits.
You need the whole screen to be visible for it to work. If grabbing the picture with a decient webcam is fast enough i'll probably move to using a few LEDs around the screen.

Cool,  I'll have to download it and try it at home tonight.

I have also wondered about the possibility of turning it around like this.  You can't go wireless this way, but it does make the positionong calculations a lot easier, just calculate the difference between the markers and the center of the camera image.

[u_rebelscum]

I think some time and effort would be better spent improving the mame code for light guns or perhaps a driver that supports raw input and not mouse emulation. I  think this is how dual lightguns are forced to work already.

I'm not sure what you mean by "raw (lightgun) input", but IMO it's the windows driver that should be doing any calibration, computation, and/or translation.  To me, that would make the info not "raw" (but that's for a different thread ).

Mame excepts joystick or mouse inputs fine for lightgun games.  The -lightgun option is just some special code to treat the mouse input as absolute, instead of the usually relative mouse info.  Since joysticks are usually absolute, mame shouldn't have problems if the lightgun is joystick.  (The reason the old-old actlab joystick lightguns don't work wel in mame is the same reason it doesn't work well for any game: calibration sucked.)

I'm not saying mame's current "lightgun" code can't be improved: it can be.  But mame doesn't need any new code to support joystick lightguns (except maybe a way to set mame's analog acceleration to 100% from the commandline, and maybe rename the "lightgun" option to "mouselightgun").  IMO, Mame shouldn't have to add code to fix bad device drivers.

[rdagger]

I had a shot at a reverse webcam lightgun tracking system.
I basically threw together the webcam and squares detection components of opencv.
http://users.tpg.com.au/mindle/camdemo.rar
Point your webcam at the screen and hit a key to take a pic. Escape quits.
You need the whole screen to be visible for it to work. If grabbing the picture with a decent webcam is fast enough I'll probably move to using a few LEDs around the screen.

I too have started work on the reverse camera approach.

[Silver]

@u_rebelscum

by raw I basically meant non-mouse (there also seems to be a rawinput structure on the win32 msdn microsft stuff, or rawhid for input from a USB HID device - think I got the term from there).

You clearly know more about the mame setup than me, it sounds more flexible than I had been told. I was thinking that there was a data-from-light-gun-->convert to mouse input-->send-to-mame-->mame converts to game-type input indriver. I though that if you had a driver that could just feed absolut co-ordinates mame could pass that directly to the emulated game.

This may be completely irrelevant  As you say its already happy with joystick input, it probably is. If you have any experience with windows drivers in this area, I suggest dropping AndyWarne an email...

[RayB]

Forgive me for not reading all posts in this thread and perhaps repeating something already said, but Happ sells the gun optic board alone: http://www.happcontrols.com/amusement/optical/960066100smd.htm

Seems to me that figuring out a driving or additional mods to that existing circuit might be another way to go...

[brandon]

I put in a vote for sensors to triangulate the guns position just like what "Jurassic Park" and "House of the Dead 3" use.  That would mean any monitor, DLP projector, TV.. any display you want.  and two player support is obviously possible since the above games were two player.  Someone like Andy or RandyT could make an interface for the sensors and have it report joystick coordinates as an HID.  Maybe a joy2mouse type program or driver could convert the joystick coordinates to absolute mouse coordinates.  I just think that using a webcam for tracking sounds way too complicated.  I could be wrong...

[u_rebelscum]

@u_rebelscum

by raw I basically meant non-mouse (there also seems to be a rawinput structure on the win32 msdn microsft stuff, or rawhid for input from a USB HID device - think I got the term from there).

Ah, RawInput. 

- It's an winXP only thing:  crashes on winME/98, doesn't work in win2k (but at least it doesn't crash).
- The device driver sets the sturcture.  Two of the three structure types are very standard "RawMouse" and "RawJoystick", which can use normal windows drivers that come with windows.
- The third is RawHID, which means the device must be a HID (Human Interface Device).  If this is used, mame must add special codes that does what a device driver normally does.

- A minor plus, it looks like with the new minGW compiler mame is now using, rawinput is now included in the headers & libraries;  Before the upgrade, it wasn't.

...I was thinking that there was a data-from-light-gun-->convert to mouse input-->send-to-mame-->mame converts to game-type input in driver. I though that if you had a driver that could just feed absolut co-ordinates mame could pass that directly to the emulated game.

Umm, since joystick and mice can pass absolute coordinates, why invent a special driver and data structure that does the same?   

Of course, if the "data-from-lightgun" is something that's not the coordinates, such as time from start of screen refresh like from a real arcade lightgun (over simplified), at some point it needs to be converted to a coordinates before mame gets it (well, unless mame starts simulating the circuits that did the convertion on the real arcades).

This may be completely irrelevant  As you say its already happy with joystick input, it probably is. If you have any experience with windows drivers in this area, I suggest dropping AndyWarne an email...

I know mame's input code okay; but write windows drivers I've never done.  I don't even have time to rewrite some features I need to to make MameAnalog+ fully functional again ATM, so learning device drivers is out for me for now.

[shmokes]

I put in a vote for sensors to triangulate the guns position just like what "Jurassic Park" and "House of the Dead 3" use.  That would mean any monitor, DLP projector, TV.. any display you want.  and two player support is obviously possible since the above games were two player.  Someone like Andy or RandyT could make an interface for the sensors and have it report joystick coordinates as an HID.  Maybe a joy2mouse type program or driver could convert the joystick coordinates to absolute mouse coordinates.  I just think that using a webcam for tracking sounds way too complicated.  I could be wrong...

IIRC Andy posted somewhere, just in the last month or so, that he IS currently looking into a product to do just this.  I think he specifically wanted to make something to work with the NAMCO guns.  I sure hope he does cos I've got a pair of Act Labs guns new in box so I'll be certian to be able to afford whatever solution he comes up with after I Ebay my Act-labs guns.  If nobody comes up with anything else at least I can fall back on my Act-labs guns.

[NoBonus]

Though this product is not a lightgun, it may give some design ideas or inspiration. It would be fun to have just to mess with though.

http://www.atruereview.com/pistol_mouse/index.php

[brandon]

Though this product is not a lightgun, it may give some design ideas or inspiration. It would be fun to have just to mess with though.

http://www.atruereview.com/pistol_mouse/index.php

sort of looks like "Lazer Tag" or was it "Photon"?    that reminds me.. I wish I still has all that stuff...

[IntruderAlert]

how about building it inside one of these babys?


http://www.studiocreations.com/howto/blasterrifle/main.html

[rdagger]

I'm leaning toward the Plager Katsumate Series-D blaster.

[IntruderAlert]

btw, as another way...

there are some very cheap crappy guns that work in mame.  Whats the chance if we hacked them into a true arcade gun, and the arcade guns original optics... that it would be better then the original?

It might make sense to use either the bio gun or the other crappy but kinda working still selling guns and fit them in a better case... with better optics.

might be an option.

yeah.. that's what I was talking about earlier
just hack an existing light gun interface to use a real arcade gun
I mentioned act labs but any cheapo device might do as well as long as you are using the real arcade gun optics and all

[NoBonus]

You know, since this is BYOAC and we all have different ideas about what the gun should be like, why not just produce the guts of the gun and let the people who buy the guts find a suitable gun to put it in.  Maybe it could even include all of the triggers and buttons, just not in a gun case.  That way if someone wanted a rifle, they could make one.  If someone wanted a .45, they could make one.  This would make the possiblities limitless.  Of course, it could be produced with a "complete gun" setup and a "gun kit" setup.  This way those who like the BYOAC model can get it, and those who want something different could just get the kit and go nuts.  I, for one, would incorporate it so the barrel mech fits over my eye and the trigger is on my temple...

[brandon]

I wonder if you could use an Analog Devices ADXL202 accelerometer?

[ErikRuud]

That is similar to the technology in the Gyro Mice.  It would sort of work, but it would not be true aiming.

[brandon]

as long as the shot goes where the barrels pointed I dont really care

http://www.analog.com/en/subCat/0,2879,764%255F801%255F0%255F%255F0%255F,00.html

They sell gyros and accelerometers.. I think if you can make a VR glove with them you can make a gun..but then again those guy were at University of California, Berkeley.  They might have more between their ears than most normal humans 

[ErikRuud]

The problem with gyro mice as guns is that they have no direct link to the screen.  They also don't allow for a "natural" gun aiming motion.  They register twisting at the wrist, better than movement of the whole arm.  An accellerometer based system would probably handle arm movements better, but I don't think that it would get confused by the user changing position.

[brandon]

yeah.. I thought about that.. the gyros and stuff would detect tilt and movement but you'd still need something to tell the PC the point of origin.. This makes hacking some Guncon's sound SO much easier now

[Lilwolf]

It would be pretty cool to be able to hack one to get two machine gun squirt guns to use them with T2 type joystick gun games... make them wireless and your would have a really cool sweet remote T2 sessions.

[brandon]

hmm.. that gives me a thought.. I have a wingman extreme Gamepad with a "sensor" button.. when I enable that it actually lets me tilt the gamepad and it translated it like analog on the X & Y joystick axis.  I could hack that tilt sensor into a gun housing thats mounted to a control panel and make a T2 positional gun without pots!.. or I think I could... sounds worth a try at least..

[dan57fina]

so 1up...do you mean to say that if we could convince the MAME development team to add guncon2/ror3 support, we could simply plug these devices in and play?

[gl.tter]

Interesting thread guys.