Greetings,

I recently purchased the iPi Mocap Pro license and am trying to resolve a scale issue with data exports.

Exporting the calibrated scene gives me cameras that are positioned on a scale of meters (according to the export file) -- the cameras are not labeled, but I'm assuming order corresponds to camera number (eg first camera in the file is camera #1, second in file is camera #2, etc).

Exporting BVH gives me a BVH file with no scale. The values definitely aren't in meters -- what are they in, inches? I'm trying to re-align the original camera with the BVH animation during playback but so far have been unable to do so.

Hi
BVH is exported with distance values in centimeters and angle values in degrees.

Thanks for your reply.

One question regarding the "Save Scene" functionality from the Scene Tab in Mocap Studio:

When I look at a particular camera in my setup in this tab, I see certain pan/tilt/roll values, for example:

Position XYZ: -2.887, 1.257, -1.115
Pan: 116.34 degrees
Tilt: -10.38 degrees
Roll: -6.92 degrees

However when I export the scene using "Save Scene", my resulting scene file contains this:

<Translation value="-2.887034 1.256659 -1.114777" />
<Rotation axis="1.00000 0.0000 0.00000" angle="1.55935500"/>
<Rotation axis="0.00000 1.0000 0.00000" angle="-64.826700"/>
<Rotation axis="0.00000 0.0000 1.00000" angle="-164.698600"/>

While the X/Y/Z coords of the exported camera match, the rotation values do not. Is this a bug or am I confusing something?

...

I am taking it you are attempting the re-alignment in an external 3D editor for a video plane?

Which program?

The transfer is a bit tricky, but really all you need to use is the scene tab to get the cam position X,Y,Z values (negative value in iPi may become positive value in editor possibly), and the Target X,Y,Z values to set depth of video plane. (easier to look at actual scene tab to get values than to use the scene file).

Set FOV of cam in editor to whatever it says in iPi, PS Eyes should be 75 degrees.

You also may have to manually rotate the video plane back a bit to match lean to camera, do this by eye easier.

Scale of skeleton should be selected as centimeters on import, or set to automatic.

Easier to pan, tilt, roll cam to match better in editor scene and use those values in script, but I don't know really what you are attempting with the script.

...

Sorry, I had a typo in my original pasting of camera export information. Below is the proper data:

Position XYZ: -2.887, 1.257, -1.115
Pan: 116.34 degrees
Tilt: -10.38 degrees
Roll: -6.92 degrees

However when I export the scene using "Save Scene", my resulting scene file contains this:

<Translation value="-2.887034 1.256659 -1.114777" />
<Rotation axis="1.00000 0.0000 0.00000" angle="155.935500"/>
<Rotation axis="0.00000 1.0000 0.00000" angle="-64.826700"/>
<Rotation axis="0.00000 0.0000 1.00000" angle="-164.698600"/>

Snapz: I am attempting to verify these numbers in both Blender and a custom-built Python program. Neither set of angles seem to reposition the camera to even roughly align with the original camera rotation (the X/Y/Z positions seem correct, however)

...

Since there is no data in a BVH file for camera positioning in scene, unless your cameras are permanently fixed in positions in the real world, how will Blender recognize these values for each imported BVH file?

Are you using a direct file path in the script to the iPi scene file to get your values to attempt this script?

I have set this up in 3DS Max before, have not tried Blender, but should be similar, but I have no idea about a custom program.

...

I'm manually positioning cameras x/y/z positions and pitch/yaw/roll in my python program and Blender to match those provided in the iPi scene export.

My concern is the iPi software GUI (the "Scene" tab that shows camera information) doesn't show matching pitch/yaw/roll values as what iPi exports in its scene file. Note that the camera X/Y/Z positions *do* match however. So either one of these sets of values is wrong (a software bug in iPi Mocap Studio) or I'm not understanding something about how this data is exported.

...

Yaw,pitch and roll values wouldn't match from iPi, that is only for iPi scene and how those cameras are set to record only the iPi recording and re-display it for tracking.

When you import into another program, your skeleton will come in square to that scene at 0,0,0 without any camera yaw, pitch, or roll needed in a new editor scene, but it is easier to manually match the plane to the skeleton and then see what the program gives as those values.

It seems Blender doesn't allow a target value to be set on the camera to set the plane depth, does it? I don't see it if so, so manual depth would have to be done just to match imported skeletons scale (height).

I found some camera parameter settings... Blender is not my editor of choice, hard to find options for each object, but if you play around with it knowing its UI better I am sure what you are trying to achieve is possible, just getting the camera in scene at proper X,Y,Z position and FOV is most important, plane can be manually set to match through camera view.
...

Both sets of values presented on the Scene tab and in .iPiScene file are Euler angles, but with different order of rotations around axes. The order on the tab is X (altitude), Z (roll), Y (azimuth). The order in a file is X (first angle), Y (second angle), Z (third angle). You may experiment a bit to find proper direction of rotation (whether you should negate some of the angles).

I cannot comment now on the reasons for this difference in view and storage. One thing is for sure - when .iPiScene format was designed there was no thought that files will be directly viewed or edited by users, thus it was not intended to be user friendly :)

...

Correct, all those values only matter within iPi itself, will not translate to an external editors scene or camera, except the X,Y,Z camera positioning and the Target planes X,Y,Z values and should be the only values needed to re-align skeleton to video, but if the plane rotation to camera is off a bit, just manually rotate it a bit better.

Then video scale to skeleton scale should maintain closely throughout the video playback.

As for making all that happen with a Python script, or in a custom application automatically, I can't comment.

I guess using the Y,P,R of the camera externally may be used for an exact (all camera) re-created scene externally, but for 1 plane to be aligned to camera should be achievable, but I am not sure the exported iPi scene file values will manipulate things they way they are needed.

I have tried a multi camera scene camera set up in 3DS Max and it was hard and I had to manually manipulate some values to get it to work with the exact real world positioning of the cameras, because some of the video planes wanted to over lap each other the way I have my cameras set up, so I didn't find an exact external scene creation possible with multi cameras, I could see only a head-on and 90 degree plane feasible for skeletal editing purposes.

...