Im using the Logitech Pro 9000s. I did try and reshoot my t pose and moved one of the cameras to try and recalibrate. I did go through the fov calibration on both camera a couple of times. Oh sorry that long image of calibration sequence was from another clip that I didn't go above my shoulders..
It looks like the roll on my camera is way to high too.
I was thinking that it may be that I am in such a small space. Do you think that a larger area would help?
Im in LA and its hard to find a big open office with no people..IF there is anyone in LA that wants to help and has a place to do it please post :)
Also do you think a 3rd camera would help?

Good question about using a larger area - not sure about that one. The calibration tutorial I posted was shot on a soundstage, so I had plenty of space, but I have shot two camera setups at my office with much less space and it worked out ok.

A 3rd camera might help, but not for calibration necessarily. Two cameras should work. One other thing that might be throwing off the calibration is that the flashlight is pretty far away from the camera during the whole clip, so it could simply be that a new calibration pass needs to be photographed with the flashlight getting much closer to the cameras, at least for part of the time, so that DMC can correctly calculate the depth of the space, especially when it is small.

If you look at the calibration clip, you can see that for the left camera, the flashlight only covers the left side of frame and never covers the screen-right space, and the right camera never moves into the screen-left space, so you're effectively not using 1/2 the screen for both cameras, and there isn't much z-depth either.

Hi,

I just checked your files. The problem is definitely related to calibration video sequence. In second camera view, you have a picture on the wall covered with glass, that reflects lots of light. The system mistakes reflections for the marker in 90% of frames.

http://www.ipisoft.com/downloads/video/misdetection.jpg


We (developers) should make the marker detection more robust in upcoming versions. For now, I recommend a workaround:

I would suggest that you re-shoot your calibration sequence. You can cover the picture on the wall with a piece of paper, or just take it away for the time of video shoot. This should help to make a good calibration video.

I do not see other problems in your videos. The T-pose is OK and everything else is OK. As soon as you manage to get good calibration, everything should work.

The problem with your calibration video is more tricky then just a simple misdetection.

I managed to get good marker detections by adjusting the background region, but still cannot get correct calibration. I will let you know as soon as we find the cause of the problem.

I did shoot another calibration video here: www.samgebhardt.com/tmp/video_2009-04-3 ... 420CAL.rar
I tried to cover most of the glass. but see that Im still getting alot of reflection
I also moved the cameras
cam 1 1.75 m
cam 1 to 2 2.3? m

Your first calibration sequence is actually more or less good. Here is project file with reconstructed marker trajectory and resulting scene:

http://www.ipisoft.com/downloads/video/ ... _CAL_5.zip

You have 2 almost identical workplaces (similarly looking tables, chairs, LCD screens) placed back-to-back in the room. This was a bit confusing and I spent some time figuring out where is left and where is right on video.

Here is motion captured you:

http://www.ipisoft.com/downloads/video/video_2009-04-30_12_46_50_i420.zip

You can get better results by using more cameras, solid-color shoes and less-loose-fitting jeans.

NICE!!! WOW this still looks great!. Im have to run but I will look at this more mon. THANKS!

ok this has been solved already, but I want to add something, I have had that problem before the best way to end up with a good camera calibration is to play with the contrast of the video once it is recorded, I also had a lot of noise on some recordings, "reflections, lights, floor reflecting the calibration object", then I grabbed the calibration video, placed it inside a video editing software and made it as dark as possible working with the contrast, then when I was basically unable to see with my naked eye the spot, I saved it and did the calibration.

It was perfect !

As I have stated before this program works VERY VERY VERY good, it just takes a bit of practice and some tricks, that I guess you learn by testing it, the last 2 weeks or so I have been super busy and I have not been able to do more testing, but the ones I have done so far with 2 cameras, I can't really complain about the results.

good call I was thinking about pulling the calibration into AE and masking out everying but the light

So thats what it was? I was getting to much noise and "distraction"?. on my calibration video?
..Im asking
Thanks for your help. im going to get a couple more cameras and give it another go Mon.