Hello all,
I am excited to have found this product. I am working on an indie game with Unreal and so far I was forced to hobble together some stock animations from different sources and hand animation sequences from still poses, which look not too good. I hope to be able to make my own animations with this system.

I have some problems understanding how this work and what the best setup is. I will buy a separate computer of course. I think I have also figured out that 8 Sony cams is a pretty good setup, better than the 2 Kinects I was thinking about first.

What I don't understand is, where the graphic card comes in and most importantly how the frame rates fit in that ipisoft gives on their graphic card comparison page.

I understand that you first record a film with your cameras and edit it, then feed it into the actual motion tracker that calculates the skeleton data. The graphic cards on the compare page vary with their framrates from 2 to 30. Now: Does that mean that the CAPTURE process witht eh recording software is only 2 or 10 or whatever the graphic card provides? Or does that mean that that many frames are later calculated per second with the motion tracker software and that the conversion will take that much longer?

I am hesitant to buy a 1000$ graphic card when all it does is to make the post processing 4 times faster. However I want a good animation sample rate. 10 samples per second or so are surely too low to look good, but I would think that simply recording a video with say 30 FPS or so will then also create an animation with 30 samples per second, regardless how long the transformation takes?

I can wait a while for each transformation but I dont want choppy animatinos in the end. So where does the graphic card performance fit in here?

Thanks

Hi

Recording process is done with that frame rate which cameras or depth sensors are capable of. There are some hardware requirements for recording so the performance is not degraded, but it's not a high-end GPU. More concerns about USB controllers/cables and storage.
GPU performance seriously comes into play only during the tracking process in Mocap Studio. So FPS you see in comparison chart is a number of frames processed per second and does not affect the speed of resulting animation. FYI, recent versions of Mocap Studio also allow to specify the target frame rate when exporting animation.

Find more info about hardware requirements for recording and processing in our docs:
http://docs.ipisoft.com/iPi_Recorder_System_Requirements_for_Multiple_PS_Eye_Cameras_Configuration
http://docs.ipisoft.com/iPi_Mocap_Studio_System_Requirements

Cool, thanks alot! That is very helpful already :)
So I will rather spend my money on the USB and camera equipment than on a overly powerful graphiccard.

One other question: Is the CPU a determining factor when capturing? I read in the links that even a pentium 4 would do to capture? That would be good news, I have a 3rd Generation I3 which I dont use anymore but I am thinking if I equip it with several USB controllers and SSD drives for capturing and then do the motion capture processing on a high end PC, would that work? Or is soch a machine too slow to capture the videos?

I'm trying to put this machine to good use since I don't use it in my daily work so much anymore

CPU usage during recording depends on a camera configuration and compression options used. The more cameras and/or more aggressive compression the more powerful CPU is required. We usually recommend 4-cores CPU for recording.
A 3d generation Core i3 has 2 cores + HyperThreading, so I suppose it can cope with moderate camera configurations like 1 Kinect v2 or 2x Kinect v1 or 4x PS Eye. Not sure about 8 PS Eyes though - maybe writing uncompressed video to SSD will be a viable option. If not, then you may try reducing frame rate to 50 or 40 if that works for you.

Thanks. I found a USB 3 card on newegg that has 4 controllers and offers a combined data transfer rate of 20Gbs (https://www.newegg.com/Product/Product.aspx?Item=N82E16812400546)
I will buy 2 of them, plus 2 SSD drives for a Raid setup. Hopefully this will work then with uncompressed recordings. I need mainly short animation sequences that I can run in a loop or short attack sequences, with any luck this might work. 40 or 50 samples per second should be good enough I assume.
Sounds really good. This way I can reuse my old computer in the garage to record all the animations and get a new one for the actual motion tracking that I can use for other stuff as well at my desk.

If you plan to use up to 8 PS Eyes then this equpment looks a bit overkill.
Single SSD drive (even not dedicated) is quite enough for recording with 8 PS Eyes, no need for a RAID unless you have other tasks for it. And it won't offload more work from CPU during recording if you meant that.
Yes, these multi-controller USB 3.0 cards from StarTech show good performance with PS Eyes and are quite useful. However, you should be able to connect at least 4 PS Eyes to onboard USB controller(s) in case there are sufficient number of available USB ports. For another 4 PS Eyes you can get one dual-controller card with 4 ports (2 ports per controller)
https://www.newegg.com/Product/Product.aspx?Item=N82E16812400547&cm_re=PEXUSB3S42V-_-12-400-547-_-Product

EDIT:
Also, be aware that Basic license supports up to 6 PS Eyes, and for 8 PS Eyes you need Pro license which is substantially more expensive. You may start with 6 PS Eyes and 3-month Basic license to get acquainted with the software. And if you see that 6 cameras are not sufficient for motions you track or you need some advanced features of Pro edition then purchase Pro license for the next period.

Oh, I didn't realize that there is a 3 month license version. I was eyeing the pro version, but yes, it might be an option.
What is the actual difference between 6 and 8 cameras? I mean the "quality" is better but how exactly does that play out? Are the animations captured more exactly? Or better fault handling? Or van greater details be captured like foot or hand position?
Of course I'm happy to save money. I will need pretty exactl head and hand capturing though since I want to also do aminations where the character manipulates objects like machines or does delicate things like tinkering. I don't need finger capturing of course but hand, food and head orientation should be pretty exact.

I read already that motion controllers might be necessary for that.

More cameras generally mean higher accuracy and few occurrences of occlusion, i.e., body parts blocking other body parts. The more iPi can see, the less it has to guess about what's going on in the video.

It also means you'll be able to capture two performers at once since there is less chance that they'll occlude each other.

The motion controller works well for head tracking. You'll need to come up with a way to attach it to the head. I use a modified hard hat; some users have had good luck with a GoPro strap. I think Snapz just uses the interpolated head tracking (without the motion controller.)

Motion controllers also work well for wrist capture. It's amazing how much more realistic the animation becomes with wrist capture data. No finger capture but Mocap Studio has a good system for adding preset hand poses and in-betweening the poses.

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

Watch a one minute excerpt on Vimeo now!

Thanks for helping out, Dennis :)
Nothing to add except for a link to the docs about motion controllers
http://docs.ipisoft.com/Motion_Controllers

Thanks a lot. I think I have enough information now to buy the necessary hardware and then I will check out the differences between 6 and 8 cameras in the try out version before I decide which version to license.