I've use iPi Studio basic with dual kinects, but its been almost a year ago. On my main system (where I run iPi Studio), calibration and mocap processing were very quick processes with the older version of iPi Studio (this system has a 3.2 Ghz i7, 16 GB ram, and GTX 580 classified). My dual kinects and iPi Recorder are configured on a secondary system in a larger room on the lower level of my house. I'm using the current version of iPi Recorder and Studio and I knew the calibration process was a little different, but I did not expect any issues.

I did not have any trouble with creating the dual kinect recordings. For the calibration recording, I moved a large card board (within recommended specs) between the two sensors for around 30 seconds. The trouble started when I tried to process these in iPi Studio. I clicked on 'New' and then identified the calibration session/recording file. I then clicked 'Calibration Project' and then iPi Studio hung at 'Detecting ground plan, please wait'. I figured maybe the calibration process was slower than I remembered. So I left to do other things. But when I returned after an hour, the 'Detecting ground plane ...' message was still there.

Does any one have any ideas on how to get past this issue?

Hello
Have you tried to open the calibration video a second time? Maybe, this was an occasional glitch?
If the issue still remains, can you provide us this video?

I should have added that I tried this several times but got the same results. I also rebooted my machine and tried it a couple of times. I'll attach a file when I get to my system (posting this from a tablet).

Please also try one of your past calibration videos, if you still have any. Also record another one, just for test. Do they suffer from the same issue?

In my last post I guess I wasn't clear again. In the several times I retried the process, I created new calibration recordings (four or five) and had the same problem with each one.

How long / large should a calibration recording / session file be? I ask because I tried to upload my 1.2 GB calibration file as an attachment here and it fails to attach.

I will try to use calibration I recorded around a year ago. In the past the calibration footage and the performance footage were recorded into one file. Can this new process use that?

From the prospect of the calibration process we recommend to record about 1 minute long video.
As for uploading - the forum is not intended place for this. You can upload to some cloud storage like DropBox or Google Drive, and post the link here. I recommend you to zip the video before uploading.


In version 1, there was simply no notion of a project kind (calibration or action). But each project corresponded to a single video, the same way as now. If you only have previous records made with v1 (AVI files), you can convert them to iPiVideo format using a built-in conversion tool in iPi Recorder v2. Old project files cannot be opened with iPi Mocap Studio v2.

Here is a link to one of the calibration files for your review.
https://drive.google.com/file/d/0BwVI2Tkh2Zk4dVN6TzY4d1dNOG8/edit?usp=sharing

Thanks for the link. I've downloaded the file and checked it.

There was no freezing issue for me, so definitely it has something related to your PC.
Do you have the latest drivers for your video card installed? Please check with the NVIDIA website.

I also have a few tips for improving calibration results, based on this video.
First, I recommend you to hold a board in the middle of its side rather than near the corner. This allows the program to identify the board more reliably.
Next, prefer doing more motions forth and back, left and right and not rotating the board largely.
Finally, it seems that this board was not perfectly flat (it's bent in the middle) - this also may confuse the calibration algorithm.

I want to thank you for your time. I really appreciate the support.
After updating my graphics card drivers to the latest level, the hanging problem went away!
Also, I will heed your advise when generating future calibration videos. Thanks again!!!

Edit:
One last thing regarding that calibration file. The 'error angle' was 1.0 and the 'average position error' was .015. I know we should try for low values. Are these consider good or bad values? What range is consider good for these?

Unfortunately, we do not yet have exact relations between these statistical values and the quality of calibration. They can be used reliably for identifying unusable results, when the values exceed reasonable limits. But there's no straight way to determine whether calibration is pretty good when the values are relatively low. Your values fall in this range of "probably good".
Usually, we estimate the result of dual-Kinect calibration by looking how well the point clouds of a calibration board from two sensors match each other, through the whole calibration video. In ideal calibration, they should be the same or very close parallel rectangles in all the frames.

BTW, you can watch this video as a sample of how one should move a board during the calibration shoot.