Hello guys

We are starting a small pre-vis studio and we will have a room that will be used for motion capture.
Everybody is saying that PS3 Eye cams are hard to setup and basically 2 kinect sensors can be setup in no time.If we are going to have one room just for mocap is it better to go with 2 kinect sensors or 6 PS3 eye cams or maybe just regular webcams (http://www.youtube.com/watch?feature=endscreen&NR=1&v=LiSYCD6fUD0-this mocap is with 4 webcams.Once we setup those cams\sensors we'll have no need to move them at all.My main question is what setup will give the best performance???
room size is 6x6 square metters

thank you for your answer

Dual kinects for sure.
That being said, since you'll be having a dedicated mocap space, and since you'll be getting at least the basic package (if not the standard), you should consider also getting the 4 or 6 ps3 cam setup. The cost is negligible and, with the 4 cam setup at least, doesn't require any significant upgrades in hardware (ie, an SSD) from what you're already likely using for your previs.
The necessity for that, however, can be evaluated once you've implemented your dual kinect setup. One of the benefits of the ps3 cams is they can capture a larger performance area and faster motions. The drawback is that conditions have to be ideal (special clothing, lighting, etc...). The kinects are FAR more forgiving in that regard.

Stay away from regular webcams, they all seem to have auto focus.

With the size of your room the 6 PS3 eye cams will work. Using DMC is a skill that is not hard to learn. You might not get instance results on the first try.

2 kinect sensors give you a quicker setup than the six camera but you lose the larger capture space of the 6 cam setup.

If you plan to have the cameras on tripods, they will get moved. A recalibration will be necessary.

If I had the space, I would get both. 300 for the knect and 400 for the 6 PS3 eye cam setup, camera and cables or 700 for both is dirt cheap. The computer to record needs to be fairly new, but the computer to render out the mocap needs a high-end graphics card. Big dollar graphics cards will cut down the render time. What takes the crappy onboard graphics card 25 minutes to render will be done in 6 minutes with a 400 dollar graphics card.

But it all comes down to your needs. Is a meter by meter good enough work space? Then the kencts. Or do you need a 3 by 3 meter work space? Then the 6 cam setup.

Dale

This is just my two bit opinion, but I would have to say the ps3 eyecam setup will be your better bet. Although the Kinect setup is fantastic, you are still limited in cap area size, and frame rate.

I use 6 PS3 cams. This setup is not hard at all to set up and calibrate. Yes, it may take a few minutes longer to setup, but once it is up, you'll never have to calibrate again.

In addition, 6 cams will allow you to cap more than one person, allow for more subtle movements, allow for a larger cap volume (stage area), and more complicated movements.

In the long run, it would be worth it to get both, but to start off I would highly recommend the Eyecam setup.

Pat

I thing I will probably get 6 Eye cams,tripods,lighting etc.

Money is not an issue,since we all have been working really really hard all year a collected enought to start our own small business. :-D...but always is better to go cheaper if you can

Just one question.I have Nvidia 570GTX Twin Frozr 3 graphics card,will that be enought to produce 'fast' results.Are you talking about SSD just because they are way faster or there are some issues doing mocap with PS eye on HDD-this question is maybe stupid,but it's better to ask than regret later. :-D

For six PS3 Eye cameras, I highly recommend using SSD for capture. In fact, I would recommend SSD for four PS3 Eye cameras or two Kinects even. A lot of times, when you hear about 'buffer problems' in these forums, it's because of slow drives. Remember, you're attempting to push huge amount of data from multiple sources through your system and it all has to be perfectly synchronized. This is quite a feat for any 'consumer grade' desktop computer. (Kudos to iPi Soft for making this even possible.)

BTW, you don't need to purchase a huge SSD. I use a 120 GB SSD for iPi Recorder data but even if you get a 40GB one, that's plenty of space to dedicate to capturing. After capturing, you may want to move the data to a regular 'work' hard disk. That's what I do.

As for PS3 Eye vs. Kinect. This has been discussed many times in these forums and everybody has a different opinion. I use both systems and, if it's an option for you, I recommend getting both. If you can choose only one, it depends on what you want to capture and how much time and effort you want to put in.

In short, the quality of two Kinects is comparable to four PS3 Eye cameras. Kinect is significantly easier to work with but PS3 Eye gives you a much larger capture space.

Six PS3 Eye cameras can give you larger capture space, more accurate coverage, and higher resolution but it takes more effort and it also requires more computer 'horsepower' to capture this much data than the other configurations.

So what do I like? Currently, I love using the dual Kinect configuration. It's simple to set up in a small room, and the results are quite good. That said, this forces me plan most my shots to have very little walking in them. I still use the six PS3 Eye system for bigger or more complicated motions but I try to save these for one big capture session because the PS3 Eye configuration requires me to move my computer to an offsite location. I really didn't mind this back when PS3 Eye was my only option but the ease of dual Kinect has me so spoiled now. :p

I think your graphics card should be ideal. Mine is an older GTX 480 and it's definitely fast enough for production work.

Hope this helps.

G.

_________________
Greenlaw
Artist/Partner - Little Green Dog | My Demo Reels (2013,) (2015,) (2017,) and (2019)

Watch a one minute excerpt on Vimeo now!

thank you

Everything you said was quite helpful :-D

Yeah, Ive used 1 kinect, 2 kinects, 4 ps3, 6 ps3, and 6 ps3 is definately the best... kinects major shortcoming is with foot contact action, 3or4 ps3 cam do a better job of this. THat said though, foot motion is always a lil slippy no matter what, so cleanup is always gonna be needed (if the feet are gonna be in shot ofc).

as for the whole SSD thing... I gotta say there seem to be better drives. Ive got an 80gb ssd (vertex2) and a 300gb Velociraptor in my machine. Some of my capture sessions have run to 2-3 mins for a single recording, and the ssd just cannot maintain that kind of datarate for that amount of time... the velociraptor kicks it way out into touch by comparision.

Also, dont go skimping on size... I did a capture session for a commercial the otehr day (with dual kinect)... a few minutes of footage in total, and it took up over 70GB of space... so be aware.

_________________
My Site

I'm sure this varies from case to case but I found that turning up CJR for the legs can solve much of this. Of course, too much CJR can make the result look unnaturally smooth and robotic, so it can be a balancing act.


I've recorded sessions that ran far longer (maybe 5 minutes or more, which I should point out is way way too long to deal with on iPi Studio's timeline, so DON'T DO IT,) and have had no problems with the SSD. Maybe it has to do with the quality of the SSD? I don't know. FWIW, I use this model in my desktop computer:

http://www.newegg.com/Product/Product.a ... 6820233125

I think by now this model been replaced by better ones but maybe the specs will give users an idea about what works.


I have to agree--get as much storage as you can afford. No matter how much space you get, it will soon become 'not enough'. :)

That said, you can reduce your data requirements significantly by disabling RGB for Kinect data or using compression for PS3 Eye. For example, for 'Happy Box' I recorded multiple takes for three characters for 17 shots, and with RGB disabled, and I didn't come anywhere near filling up my SSD. And this includes the entire project shot twice: Happy Box was first shot with single Kinect, and later it was completely reshot with dual Kinects when this option became available. That was a lot of raw mocap data. :p

As iPi Soft explained in earlier posts, using compression does not affect mocap quality very much, if at all. In fact, maintaining a constant framerate is far more important than compression or resolution even.

G.

_________________
Greenlaw
Artist/Partner - Little Green Dog | My Demo Reels (2013,) (2015,) (2017,) and (2019)

Watch a one minute excerpt on Vimeo now!

If memory is a problem you can record one action then save BVH file,delete video,record the rest of the action and blend it all together in MotionBuilder.Even if it's not a memory issue I think it's better to record more than one video...for example if guy need's to jump,than do a back flip,than hand stand,than front flip etc...Proffesional can do it all at once but the motion won't be that fluid.
As for the foot sliding and things like that-I learned never to rely just on mocap,some things just need to be fixed with regular key frames :-D