I am trying to get a successful track using 6 ps3 eyes. During calibration I find that about three of the camera positions are calculated correctly, while the others end up rotated or under the ground plane. Even I dont use the auto detect camera option, I get weirdness in cameras changing position and rotation. Any help in what I might be doing wrong would be greatly appreciated. Thanks in advance :)

I felt your pain and i just want you to know i had the same problem but with only 3 PSeyes..Unfortunately i have nothing to offer but i wanted you to know your not alone with this issue..Sadly i lost the fight in this one and went with two Kinects instead..I am interested to in how to fix that problem still

Good luck =)

Hi,

Assuming the captured data is good to begin with, the Auto Detect Positions option should work well and you should never have to manually move your cameras. If fact, it's highly recommended that you use Auto Detect. At least, that's been my experience.

There are many things that can cause calibration to fail (poor lighting, too many reflections, incorrect lens/FOV setting, inappropriate clothing or background, etc.,) and most are easily prevented if you're aware of them. If you could post an example of your calibration video somewhere, maybe we can try to figure out where the problem is.

G.

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

Watch a one minute excerpt on Vimeo now!

Hey

To add onto Greenlaw's post, make sure your PSeyes are set to wide lens, if you twist the lens as far to the right as possible it should be fine.

Any other problems I have had with the auto detect being off is when certain cameras cant see any of the ground points that have been set. I had one instance where one of my cameras couldn't see the floor at all (it was a test), the other 2 cameras were fine but that third camera had the floor plane going through my chest. I'd make sure you don't have any sporadic camera angles and try to keep those floor detection points in view.

Oli

I checked that my cameras were wide lens. Did multiple tests, with auto detect and without, without moving cameras and with moving them. I would say my capture environment is probably less than ideal. The issue seems to be consistent tho generally, the last 2 or 3 cameras in the scene calculate ending up way below the ground plane. Here is a link to my initial test, welcome to see if it calibrates for you.

https://dl.dropbox.com/u/12853071/pseye ... 3.iPiVideo

Here are some more tests, we even went as far as to try our best to leave the light visible to all cameras as much as possible:

Here is Neal having a try at keeping it visible as much as possible:

https://dl.dropbox.com/u/12853071/pseye ... 5.iPiVideo

and Karen:

https://dl.dropbox.com/u/12853071/pseye ... 5.iPiVideo

Let me know if you guys can get a successful calibration.

In this test we changed environment and tried using only 4 cameras, which seemed to give worse results. Although I can see that the light is better isolated in the footage:

https://dl.dropbox.com/u/12853071/pseye ... 6.iPiVideo

Hi,

Downloading the files now. I'll let you know if I have any ideas about this tonight. :)

BTW, the third link (Karen) seems to be broken.

G.

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

Watch a one minute excerpt on Vimeo now!

That's weird: I can download the first two files fine by clicking the link but for the next two files I get a webpage full of text. I'll try right-clicking.

G.

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

Watch a one minute excerpt on Vimeo now!

Hi again,

So far I've only looked at the first video and I did notice something weird immediately: In Studio I see a partial 'twin' of the performer standing in the center of the calibration space the whole time. This is probably what's throwing off Studio because this error is making the video untrackable.

Next I looked at the video in Recorder and I can see the problem when choosing Hide Background. My guess is that the clean Background video was improperly recorded, causing data corruption as Studio attempts to remove the background. When recording a clean Background segment, you really need to be completely out of frame for this feature to work properly because this segment is basically used as a difference matte.

Another bit of strangeness: the video is playing much faster than normal speed in Studio and Recorder. Was this video processed in another program? If not, that's curious because my own videos play back at a normal frame rate.

It looks like I have all the videos downloaded now. I'll let you know if I see anything else that could be problematic.

BTW, if any of this data is good, I will need to know the exact the height of any of the cameras to calibrate. I also need to know the exact height of the performer. Without this information, it may be difficult to track the data accurately.

G.

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

Watch a one minute excerpt on Vimeo now!