I just looked at the next three videos.

The first two videos have the same 'speed up' problem but the background seems to be good. I'll try calibrating these later and see if the unusual speed has any affect--my guess is that some of the extreme 'jumps' will throw things off. Of course, I will need the camera heights to complete the process.

The final video clip plays back at normal speed in Recorder but it seems very dark. That should be okay as long as the Maglite is visible; unfortunately, the light does blink out several times during the session. Sometimes, the light is gone from three cameras for a long period. Is the light in candle mode as it should be? Even if it is, it's best to avoid stepping in front of the light or otherwise letting the light become occluded from any of the cameras. In theory, not all the cameras need to see the light the whole time, but you really want to keep the light visible as much as possible. If the light is not visible to all of the cameras most of the time, you can expect calibration to fail.

If you check my old calibration video, you can see how I position myself to avoid blocking four cameras from the light. Admittedly this is harder to do with six cameras in full circle but it's important to crouch below the light and be aware of where the cameras are during the recording--a poorly recorded calibration video can ruin an entire day of shooting. Also note that it was not necessary for me to darken the room to get good calibration results.

Tip: when I record for calibration, I check the results immediately and decide if I need to reshoot. The new Recorder makes it check quality. Also, I like to shoot a second calibration video after a mocap session is completed, just in case the cameras have been 'bumped' or otherwise moved. Sometimes this 'backup plan' can salvage a bunch of shots.

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!

Hey Thanks for checking it out. I looked at your video, and think your capture environment is really ideal... Unfortunately we need to capture large movements. With the amount of times we tried, I would have thought we would have had a successful calibration... To me it the pseye seems a bit finicky at this point. We dont currently have a space we can dedicate only to mocap, by painting the walls and what not. In the short term we captured some of what we needed Using 2 kinnects. We would still like to do more, since using 2 kinnects the capture area is somewhat minimal. Hopefully we can get it sorted out.

Our 'stage' was originally set up for green screen shoots and it was convenient for us to use it for motion capture, but it's not necessary to have a completely featureless background to get good mocap data. You might want to check out 1k0's behind the scenes video for his Danse Kabyle trailer. His examples show how much flexibility you actually have when using iPi DMC.

That said, one of the reasons I prefer dual Kinect right now over PS3 Eye is because you are correct: the PS3 System is a bit more finicky to use. By comparison Dual Kinect is much more 'plug-and-play'. Dual Kinect may have a smaller capture space but that also means you don't need a big space to use it. (We currently do all our capture in our living room, which is much smaller than our 'stage' in the garage.)

Hope this info helps. Feel free to keep asking questions. :)

G.

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

Watch a one minute excerpt on Vimeo now!

Hi, dimensions3
We've also looked through the videos you've provided. The speed-up problem with first videos has its roots in bad frame drops statistics during recording. The real frame rate of those videos is about 20fps for first (and that you've send us via email) and 40fps for second, and huge number of frame drops. You should pay attention to bad statistics while recording (when statistics box title turns red). We are not sure what was the cause of such bad situation. May be because you have connected more than two cameras to a single USB controller. Also Neal's videos are in QVGA mode, and in this mode PS Eyes provide raw data in different format than in VGA, and we have not super-optimized processing of this format as was done for more common VGA mode.
The last video's statistics are excellent (4 cams), but when going to calibration in Studio, the marker detection algorythm reports more than 20% of misdetections in all cameras (this value is displayed on the Scene tab for currently selected camera). This is due to occlusion of marker by the person's body. For good calibration results, this number should be less than 10%. To minimize it, the person should hold the marker distant from the body. When approaching a boundary of the capture area, it is better to hold person's body outside than the marker, so it is visible to most of the cameras.
Nevertheless, I got some promising calibration results. What I did before calibration is:

• turned off auto detection of camera positions
• reset to full circle reverse order
• adjusted cameras' directions with Pan/Tilt tools to match opposite camera's model with it's image

The project file with calibration results is attached (no scaling, because don't know cameras' heights), as well as initial approximation of cameras' positions.

PS. Capture statistics are written to iPiVideo file. There will be a viewer for that in the next release of iPi Recorder. So one could evaluate how good is the record just after the shooting looking at statistical values.