Earlier I mentioned that the right arm seemed to be 'stiffer' than the left arm. Here's a good example of where the right arm does not want to do what the left arm is doing:


This is probably related to the shoulder issue--it looks like the right shoulder is rotating forward when it should be rotating backward.

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

Watch a one minute excerpt on Vimeo now!

Except for the odd right shoulder issue, which I had to keep tweaking even when the motion was quite simple like raising and lowering arms, the software tracked the motion very well.

One problem I ran into yesterday was that I have no motion motion on the left hand PS Move Controller in any of the tests--I'm not sure what happened here but I guess it's important that you check that the device is indeed recording before you shoot. (Both Move devices were working ealier.) Would it be possible to have an indicator on the Record page to show that data is coming in for the Move and Motion Plus devices? Thanks.

Just to be sure, I'm going to try removing the software completely and reinstalling it, and then do a fresh round of tests this morning.

G.

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

Watch a one minute excerpt on Vimeo now!

I think one of my Kinect for XBox has failed--iPi keeps reporting that there is no power to the device, even though the light is on. When I look closely at the physical device, it doesn't look like the IR part is working though.

It's probably not a coincidence that this is the same Kinect device that was recording the right side of my body. I'm wondering if the data I had been recording recently was incomplete, resulting in the strange tracking results.

Luckily, we have a Kinect for Windows device here that we've been using for 3D scanning--I'll swap it and see if this works better with iPi DMC for my right shoulder and spine problems.

G.

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

Watch a one minute excerpt on Vimeo now!

Yes, I'm pretty sure the second Kinect for XBox is broken now--the same power supply and USB connection work fine with the other Kinect for XBox and the Kinect for Windows devices in Recorder, but the second XBox device fails to work with Recorder.

Will redo my tests today using the replacement.

G.

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

Watch a one minute excerpt on Vimeo now!

I had no idea that using Kinect for Windows gives you many more options for tweaking the RGB image.

It's a dreary rainy day here in Los Angeles and our room is looking quite dark on the Kinect for Xbox side of the iPi Recorder panel. On the Kienct for Windows side, however, everything is very bright and visible--even without tweaking.

Now I'm thinking we should order a second K4W device. But for today, we're bringing in a softbox to light the room.

Just to be sure, the latest implementation of head tracking does utilize RGB for assistance, does it not? If not, I won't bother with the softbox.

G.

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

Watch a one minute excerpt on Vimeo now!

No, the last version still does not use RGB data for depth sensors (and head Y rotations are not tracked for them).

Great running updates on this Greenlaw, its good to see you document this stuff for other users.

How is the Windows Kinect for 3d scanning? I have tried the xbox Kinect with reconstruct me and it is very obvious what the limitations are.

Oli

Here's something that may interest users.

As mentioned above, the RGB for Kinect for XBox was much darker than the RGB data for Kinect for Windows, so we decided to bring in a softbox to light the room (just in case room lighting has effect on the new Head Tracking system.) The result was surprising.


The softbox naturally improved the quality of the RGB data but it degraded the quality of the depth data.

This is apparently caused by the light from the softbox relecting on shiny surfaces in the room (our floor is polished wood, there are framed photos on the wall, and over on the left are many books with shiny spines,) which is apparently interfering with the IR from the Kinect sensors.

Obviously, we don't want to use the softbox for this shoot. Just one more thing to keep in mind when setting up for a shoot. :)

We decided to order a second Kinect for Windows, which should arrive tomorrow. This is going to be our last equipment purchase for our current Brudders short film but fortunately, except for mocap and lipsyncing, we're getting close to finishing this project.

G.

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

Watch a one minute excerpt on Vimeo now!

Thank you for the info! I can safely remove the softbox then. :)

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

Watch a one minute excerpt on Vimeo now!

Thanks. I'm glad this is interesting. If anything, I find documenting progres in a public forum can be very helpful for my own research and education. :)


We're using ReconstructMe and Agisoft PhotoScan. We got the idea to try this ourselves after reading 1K0's blog. We haven't spent a lot of time with either program yet because we're busy trying to finish our current movie. Here's what we know from brief testing so far.

ReconstructMe is interesting because of the immediate feedback and results. The detail is not hi-def and you'll want to retopologize the result, but the result really is not bad for a base to sculpt from. (Like, say, in ZBrush or 3D Coat.) IMO, the main issue is lack of texture data.

PhotoScan is much more capable but also requires a lot more work. It does not use Kinect--you need to capture a series of photographs of your subject, and the software will reconstruct the object from the photos with a high-res texture applied. I found that the most time-consuming part is setting up masks for the object, so you may want to do your shoot in an uncluttered room.

BTW, here's our scanning 'rig', inspired by what we saw on 1K0's blog:


I'll post more info in our blog whenever we have time to actually use this tech. For now, I want to stay focused on iPi DMC 2.0 so we can finish our movie. (Besides, it's kind of OT for this forum.) :)

FYI, I'm thinking about starting a section about the new Brudders movie at our website in a couple of weeks, which will feature more general production info. I've been holding off because there are certain aspects of the project I can't talk about yet but I expect to have clearance to talk more openly before too long.

In the meantime, enjoy my iPi progress here. :)

G.

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

Watch a one minute excerpt on Vimeo now!