|
Hi
Hopefully, I can get a bit of insight from fellow users and the guys at IPI.
Apologies it's a long post, but I'll detail out as much as I can and that might help narrow the issue down - or make somebody have a 'euraka' moment :)
I have a major problem that's slowed me down now for 2 (constant) days of set-up, calibrate, 'hand fix calibs', run motion test (etc) - then repeat... and I've resorted to the forum for advice.
Issue: Using the latest build on one Win8 system (I'm running a 30 day trial to test) or the last build I have (my licence) on a completely different Win7 machine, I am getting cameras stacked on each other... in either calibration method (180 or 90).
I initially missed this and thought it was set-up error, but I've broken down and rebuilt the setup so many times now (with the same result), that I'm stumped.
Essentially, on the latest build or the last build I had, on two different machines, no matter where I put the cameras, the calibration always results in cam 1 & 2 being stacked on each other.
This is slowing me down, as I'm trying to build a clean set-up for some important motion data I need to create, but, as I have to keep making a calibration match myself (ie manually spinning and adjusting a camera to approximate where it should be in terms of the volume), I'm sort of chasing my tail, rather than perfecting a decent capture volume, I'm fighting the fact that IPI won't recognize any of the second camera positions. So, I am rebuilding again and again, trying to get something that auto-detects 2 cameras properly (so I can then refine my capture setup).
The problem is weird, as the setup worked previously, and I've been battling this since Monday evening :(
Note: I can, as I've said, manually fix the calib (by hand) and use this to apply to captures, but I'm getting mixed results here... one really good (with the best walking I've had) but sadly, that was an earlier test and the position of the cameras and volume has changed since then and most others exhibit twitchy shoulders and the odd 'snappy leg' issue that I've seen quite a bit on older data. I'm sure I can get an optimal volume, but the fact that I am hand 'hacking' each calib, is so hit and miss... and incredibly time consuming!
Background: I've owned IPI Standard Edition for a while and last used it in production back in 2012 (although I've been re-using some of the capture from it for a while). I had a set-up (dual kinects) that worked and a good calibration for 180 (that worked best for the motions I needed).
The version I'm using is v2.1.2.19 (recorder) and V2.2.1.145 (studio) - however, this was the last update I installed but was NOT the version I captured my last working data from, that was a late 2011 version. I do remember that there were issues with stacked cameras back then, but I'd presumed the latest version would fix this.
So, I duplicated my set up (I work in the industry, so I'm used to having to TD once in a while and can be very methodical if need be), the volume is in a store room and everything was pretty much as it was (I had a bit of dusting to do, LOL!). However, this first test failed, and I subsequently went on to rebuild the space another 6 or 7 times, before I realized that the cameras where stacked.
I then thought this was the old version I was working on, so I grabbed a 30 day trial, and installed on a fresh machine (a duplicate of the laptop I use for mo-cap), as I have a few of these I use for render nodes (which means they're very clean systems - no bloatware). The fresh install (with latest MS drivers etc), gave me the exact same results... stacked cameras.
To make sure data was unique to each system (ie old and new) I captured different files for the old/new (but in the same set-up) as, obviously, the old system has older drivers etc. Again, the same issue.
So, I thought that it could be cables or the Kinects, or connection issues, or streaming etc. I've ruled each one out by testing/swapping (ie the Kinects are fine, each works on it's own - or in dual step up, albeit with stacked data). cables are fine, streaming for recording is fine. The laptops have 2 USB hubs, a USB2/3 and a USB3. Interestingly, the USB2/3 hub can handle the two connects (I normally plug into opposite hubs), so I may look at getting a 3rd Kinect at some point and seeing if that works... so, I've ruled out hardware. In terms of drivers, the old system is 1.5 and the latest IPI is running the version it wanted to download (1.8?).
So, any insights or thoughts? I'm about to rebuild the set-up again now (after some food :), but the extra hours and hours of testing/calibration due to the stacked cameras is painful.
One thing, in terms of a major feature request, in any update or next version, would it b possible to have a numeric input on the cameras, so I could simply rotate and re-position numerically, or some way of applying off-sets or a 'fix' mode that spins the data correctly? This would have greatly helped with my current problem of hand fixing the calib... it would, at least, have made things simpler.
|
