Hi all. I was tweaking IpiSoft for a week in 8 cameras configuration. I calibrated my system in auto-height mode, so cameras just found themselves in space without my intervention(i just scaled those heights). I have 4 cams with ideal height measurement and 4 other with 1,2,2 and 4 cm up&down accuracy error. I have less than 2% misdetects for glowing marker. I left this calibration as it is and just tried with variations of motions. All is tracking well but arms. They just ALWAYS gonna stick with body, even when i keep them separate from it. It`s frustrating for a bit because it happens very often. It happens even when i dressed in Dark T-shirt over longsleeve shirt. So now it`s a main problem i deal with.(I`ll attach screenshot with "bulbed" arm in itself).

So i have a few questions(that can be stupid for a bit, but...)
1. Is it a calibration problem?(I`ll attach screenshot of my calib result) if it`s a calib problem:
1.1 Should i use my "Cage" in calibration on ALL area i have.Because i thought that calibration is needed just for refinement of cameras themselves in space and relative to other and doesn`t matter where i did this cage and size of this cage, but does matter was it done clear, with floor touches?
1.2 Does it matter, did i calibration with daylight(which is from those 2 windows and it`s ambient enough) and then i done my actions with ambient fluorescent lighting at night?
1.2 in 60hz mode i have flickering cause lights are 50hz so can it be an additional problem? Now i`m recording at 50 hz and there`s no difference between cameras and lights frequences.
1.3 i used a LED lamp for calibration, not Maglite, so it have a bigger size of marker and easier to capture this light. Can this factor be significant? Should i search for smaller light marker?
1.4 Can be those height mistakes(from 1 to 4 cm) on 4 cams be a cause of capture problems too? I read read for a lot, that calibration of 4 and more cameras is pretty painful because of height misdetection(even in few meters). So i was really proud that i`ve got these results(from 1 to 4 cm error,as i said). So can this proud be false and height of cameras should be THE SAME as real or it can have inaccuracy like my calibration have?
1.5 Should the light be as bright as it can be? Or we need very soft lighting for best results? Now i have typical office boxes fluorescent lighting at night and brighter daylight from those 2 windows as you can see at picture. I read from vmaslov in another thread that LED lighting is more relevant but i can`t afford it for now.
1.5 Can a positioning of cameras be a cause? At start i wanted to locate my cams just in corners with view at center of room(6.20x6.40 and 4m height) but then decided to locate them as you can see on picture(attached it too).

2. Or it just my clothing problem? I read a lot about clothing on forums and decided that dark shoes, jeans, deep dark T-short over light long-sleeve shirt is better namely for arm tracking, which i can`t deal with.

3. Can placement of 2 additional cams(9th and 10th) on ceiling(keep in mind that it 4m ceiling) looking down on the floor be a solution for those problematic "armSticking" cases? Because it should see a gap between arm and body(so in case that my floor is light gray at least those 2 cams on ceiling should see the gray material between my dark(or light, depends on whether i wear T short over long-sleeve shirt)) arms and dark body?

So these questions are very important for me. Now i`m just trying to know all factors that REALLY influences at my recording quality. I`m asking not for just my particular case "armSticking" but for all factors because it`s important for me to get the best i can. Will appreciate any help and criticism on these questions. Thanks.

Hi

Problems with calibration are usually revealed by misalignment of actor in different cameras. When positions of cameras are incorrect, you simply cannot match the model to an actor image in all cameras simultaneously.

I suppose your problems with arms are caused by clothing. Specifically, long sleeves are very close to walls color. Try to wear something more distinct - like green or red. Also, having a print on a T-shirt is not recommended as the program expects solid color.

Also the lighting may be a problem. I see considerable shadows on long sleeves. Try to record with covered windows and ceiling lighting. You can also try to mimic your lighting by adjusting the light source position and intensity on the Scene tab.

If you wish you can provide us one of the action projects (with video) so we can have a more detailed look and possibly give you more recommendations. Use some file-sharing service like Google Drive or Dropbox.

Thanks for a quick answer.

One moment - now i`m using Basic version with 8 cams just to test this setup in life for some complex motions because i`m going to buy pro verion for 3 years. So should i send ipiLocked format or just IpiRecorder video? and should i send Calibration as project or as XML?


And what do you think about glowing marker size and size of area which i covered by lighting marker?

For action, iPiLocked + iPiVideo is enough. Camera positions are stored in a project file.
If you also want me to check the calibration video as well, then add iPiCalib + its iPiVideo.

From what I can see on screenshots, the marker size is OK. It was PS Move, I suppose? The volume covered by the marker seems a bit small. For better accuracy of the calibration a maximum volume visible to ALL cameras simultaneously should be covered.

BTW, I've forgot to mention that PS Moves are not required to glow when recording an action. And may even make some confusion for the tracking process when glowing.

So here is all needed files.
Action:
Video
https://www.dropbox.com/s/accvvtdrxvrb5 ... Video?dl=0
Project
https://www.dropbox.com/s/6r6lwanbyhajt ... ocked?dl=0
Calibration
Project
https://www.dropbox.com/s/iir8pq91wwm8d ... Calib?dl=0
Video
https://www.dropbox.com/s/mt2r4mrn38szq ... Video?dl=0

Actions are funny, i know, but it`s my goal - try all unusual motions and to know system`s possibilities.
And i don`t use PS move in calibration. It`s a LED lamp(as i mentioned in 1.3 in first post) that i consolidated with stick to avoid occlusions by myself.
I enabled lighting just to know order of controllers in hands, but if it can confuse system for a bit - i`ll just keep in mind which controller i keep in particular hand. Thanks for advise.
And please - let me know about all mistakes i`ve done. Thank you very much for support.

Also i was inspired by this video:
https://www.youtube.com/watch?v=1-jmUafH0Gs
This guy said that he didn`t use any jitter removal. So i`m wondering - such a clear tracking is a result of 100% accurate calibration(especially heights of cams)? Because i also use those 8 cameras and i still need jitter removal.
And in this video he wars white t-shirt with white walls, i just amazed.

No matter how fine is the calibration, jitter is an inevitable evil of the tracking method used by Mocap Studio. I doubt he did not use JR - maybe lost a bit of memory or confused the terms :) Or at least used high values for trajectory smoothing.

Distinct color is less important for the body than for the limbs because the latter are much easier to get lost by the tracker, especially the arms. However, selection of improper body color may lead to problems in tracking of the spine segments.

Heh:)
I made few tests today and i think i wear best clothing that i can afford.
You can see it on tests. What you think about it?
So my problem isn`t solved.
And what about those videos that i attached earlier?

I think it`s because i have not enough light in my room and have some tracking forward light from window. Also i think i have low rendering color index of my lamps and too small value of light. But i still waiting what you can say about calibration - mb i have really bad calibration despite the fact that it`s perfect...

I've looked on your videos.
The calibration seems quite OK, should not worry about it. Congrats!
Action suffers from the combination of not so good lighting and not so good clothes. With more bright lighting you'd get better tracking with these clothes. With more distinct colors you'd get better tracking with this lighting.
However, the tracking is not that bad as you think. Given your conditions, some tracking errors are expected. I could easily fix the error with the arm you've stopped the tracking on. Just tracked forward several frames after the error has occurred. Then put the arm to match the video with IK tool (approximately). Then Refit Pose and Track Backward did the trick.
Also, I suggest you to experiment a bit with tracking settings. For example, when doing a squat, a Flexible Spine option gives much better results. Enable the head tracking if the neck bends substantially during the motion. You can vary settings for different segments to maintain speed/quality balance.

About the new clothes.
+ Short sleeves are above elbow and can be correctly presented by the model in program
+ Solid colors distinct from environment
- Due to directional lighting all colors look the same in the cameras which are opposite the windows, which leads to tracking errors
- Hood is bad as it's not considered in actor's model