www.ssontech.com

[ericsh6]

I figured out how to calibrate 2 cameras and capture face markers and export to Blender. But I want to subtract the head movement from the face trackers. How is this done?

[ericsh6]

F.Y.I.

I think I can subtract out the rigid head movement in Blender but instead, I have decided to build a face capture helmet. I can get perfect face marker tracks with two cameras on tripods but I am concerned that the helmet mounted cameras will move relative to the helmet and mess up calibration. The only way to know is to try it out.

[GraphicsKid]

Good luck man! Let us know how it goes.

[ericsh6]

Another update.

I bought two Muvi Veho micro cameras and built a helmet with a brace that holds the two cameras about 7 inches in front of my face. The helmet looks like it will work but the cameras video doesn't sync. I read that the supplied micro SD cards with the cameras are slow and drop frames so I bought some class 10 replacement cards. The cameras work better but still for a 3 minute long video, the footage will be about 5 frames out of sync from beginning to end. I haven't tried tracking markers on my face yet but I imagine there will be problems with eye blinks near the last 1/4 of the video because they occur withing 5 frames. The Muvi Veho looked like the best low light, 30fps, light weight micro cameras I could find. Does anyone know how to fix the sync issue or of a better micro camera? (I prefer micro cameras with sd cards so I can capture without wires).

[ericsh6]

Update:

I built the helmet from a $10 hard hat with a tightening adjustment wheel on the back. I made two 1/2 inch aluminum tube braces and aluminum welded them to a 1/8 inch plate and hot glued them to the helmet. I clipped the micro cameras to the plate and bent the plate until the cameras pointed at my face. It was easy to start both cameras recording. I went in to my studio room which has fluorescent lights overhead (bad for light banding on CMOS cameras but they are cheap and not hot). The helmet worked great and kept the cameras trained on my face even when I made fairly sudden movements. But the micro cameras had very poor resolution and were still a few frames out of sync.

So I shelved the helmet and got out my two point and shoot cameras and put them on a tripod about 2 feet apart and about 6 feet from my face and sat on a couch with my head firmly seated on the back of the couch. The point and shoot cameras were in perfect sync, 640x480 resolution and clear resolution. I was easily able to track the face markers and track the calibration thingy I made. The only thing is that I may have subconsciously restricted my facial expressions as I was trying to keep my head static.

I still like the helmet idea because I can record body markers at the same time as face with additional cameras on tripods. But I need to find better cameras first.

I have to say the Syntheyes is very well suited for marker tracking and does a very good job and isn't too hard to use once you do it a few times. I was able to track 19 face trackers including eye blinks on 5 minutes of straight talking and have 3D points to apply to my character in about 4 hours of labor.

[GraphicsKid]

Are they out of sync consistently throughout the shot? I mean... are they ALWAYS 5 frames apart, or does it vary?

If you just start recording, blink a light in front of the cameras, then just sync it up in post.

[ericsh6]

It looks like the 1 frame per 30 seconds is out of sync so the first 30 seconds are good but then gets worse with time. But I think the resolution issue is much worse than the sync issue with these micro cameras.