...

Dance of Kesha song re-tracked with new GTX 970 video card.

Tracking was done in Low Res setting, all boxes checked, shoulders by video, and very flex spine.

Tracked: 1.85 f/s @ .54 s/f (5 PS Eye set up) Low Res option worked just as well as High Res at faster f/s rate. Refine Backward ran even faster: 1.95 f/s @ .46 s/f

I ran tracking, refine backward, Move data and JR in the same time it took to just track the video with the 750 Ti.

The less time it takes for the actor to align with the video, the smoother the outcome and the 970 was able to keep it right in alignment at that speed, very impressed.

Moves used on the hands, it is visible if watched where I was having issues with them before when this was recorded where data would go in and out of proper positioning.

No head controller used, just head tracking box ON.

The results are just tracked, refined backwards, Moves data applied and JR run with the TF set on 5, I set it this way due to seeing better results with this new VC.

I also used 0 on the torso, 1 on the arms and legs, and 5 on the head when running JR. Recording TF set to 5.

The tracking was very good to me and the feet were more stable with less hip push/lift with the recent update.

No manual editing was done with iPi translation tools, just the auto-fix methods were applied as stated above, I am not trying to show a flawless tracking, just the improved processing speed and accuracy without editing, hand issue is easy to repair with an adjustment layer, although if the Moves data was right, I would have very little clean up to do, just loop it better really.

Any frame drop blips you notice in the video are from the screen recorder, the animation is very smooth.

Link: https://youtu.be/IKb3p5nbYy4 ... View in HD Version processed with 970 FTW+

Link: http://youtu.be/lcxrqNhAsm4 .. Version processed with 750 Ti (Lesser quality recording)

...

these are all great!:D

Less cameras? I use 6 PS3 cameras. Wish I could use 16:)

...

My studio area of 3m x 7m will only handle 6 cameras and get a 3m x 3m volume, I think anymore would be a redundancy and cause longer processing times, but I also use the Basic version that's maxed at 6 PS Eyes anyway.

If I had a larger area, I can see where I would need more cameras, but would still try to use the minimum needed, but that would also require the PRO version.

I did add a 6th, but really achieved no better results, just added a bit to the processing time, I don't know how 16 cameras would process with my machine config. as it is now, I think I would probably need to upgrade to a GTX 980 Ti or even dual 980's to maintain a decent processing speed.

With the GTX 970 FTW+ (O.C. core to +75) and 5 cameras, I can now process in Low Res setting with all options ON, shoulders by video and very flex spine at 2.15+ f/s @ 47 s/f with no more tracking errors than in Hi Res and Refine Backwards running at 2.35+ f/s @ .42 s/f, JR runs at 6.7 f/s @ .13 s/f

Right now I am very pleased with my set up and the results I get on faster steady motion actions, especially since I am getting much more accurate hand Moves tracking.

Slower motions tracking, or more subtle motions are less stable, after all processing there are still glitch areas in the animations, but nothing that can't be manually corrected right in iPi Studio first before exporting, then refined in another 3D editor.

...

I always felt that that with more cameras you have more information for the cameras to track. Therefore, less to process in ipistudio when obstructed body parts can't be seen by the PS3 cameras. more cameras, less room for error. 6 certainly gets the job done. Not sure I'll be getting any new hardware to allow for more, but clearly it works well as you have things so why change:) good stuff!

this is one time I actually had 8 cameras for a few seconds:
http://adamansorge.com/experiments//moc ... a_loop.gif

never got 8 again after that:/

...

I only record single performer actions, so yes, I would guarantee if recording 2/3 performers in close proximity actions, more cameras would be needed.

I have my cameras positioned in a way that works well for what I need, the performer would have to actually sit on their feet or hands in a standard action or dance, but it has happened, especially on ground actions and breakdance moves, but usually limited enough that I can stop tracking and refit the actor and carry on, and manually fix those areas later right inside iPi Studio, which is a good feature iPi has, among others.

I won't be able to change much anyway due to the cost of iPi PRO now and my studio area, maybe when I can secure a larger studio I will try the PRO and add a few more cameras.

For the Basic cost it works well with PS Eyes recording at 60 fps, I just try to get performers to understand occlusion and try to limit it where possible.

Also, if you are doing co-op animations that aren't in real close proximity, or the cams can view each performer fairly well throughout, you can track 2 performers with Basic, (which even PRO doesn't process both performers actions at the same time), it tracks one at a time and exports them as individual files anyway, so you can do the same in Basic just by tracking the video twice, once for each performer and don't select re-center coordinates on export box and it works fine on import into a 3D editor.

Thanks for your positive comments though, I hope you are getting the results you need also from iPi.

...