Experiments

How we did a few hundred 3D scans of people at FITC Toronto

How we did a few hundred 3D scans of people at FITC Toronto

Since December every week, usually on Fridays, we’ve invited people to come by our office to be 3D scanned. It was just a convenient excuse to have people by that we hadn’t seen in a while. I would joke with them that it was my way of doing sales and R&D at the same time. Close to the end of March we had Shawn Pucknell the Director of FITC in to be scanned. We posted his scan online a week or so later and as a result he asked if we could so something like that at FITC Toronto.

Of course we said yes.

We had a problem though, our existing workflow for 3D scanning was too slow.

The scan and processing to get a high-poly mesh with vertex colours with a Microsoft Kinect and Kscan3D could take 30-45 minutes. In addition to that we needed to make a low poly version, do UV Mapping, texture baking, and add a base with other 3D software. Which could take an additional 15-20 minutes. An hour of effort per scan was just too much time to be practical.

Given that at FITC there would be hundreds of people, most of them wanting be scanned in the break between presentations we needed a system that would allow us to do a scan every minute or two.

Overcoming technical challenges like this is, well… what we do. Every project we get involved in poses some risk in one way or another. The fun ones, the ones we remember most fondly (at least once they are completed) are the ones that challenge our abilities and push the boundary of what we thought was possible a little farther back. Projects exactly like this one.

Initially I wanted to build a photogrammetric scanner with an array of Raspberry PIs and PI Cameras. By synchronizing all of the PIs to take a snapshot at the sametime we could capture the data necessary for the model in less than a second, and then process it at a later time. With 70 PIs rigged up all around the subject the quality of the 3D model would have been phenomenal. Unfortunately when we worked out the production effort it ended up at between 30 to 45 days to construct, develop and test the installation.

Another 3D Scan at FITC

Given that FITC was only three weeks away at that time this wasn’t a viable option, so it instead we upgraded our Microsoft Kinect/KScan 3D setup.

We built a motorized turntable with variable speed control, and a tower to hold four Kinects stacked vertically. By turning the subject through one full rotation while capturing data from the tower we were able do a scan in less than two minutes and that includes getting the person on and off the turntable. By holding off on the processing of the models until later we were able capture over 331 scans at FITC.

Our new and improved set up all also decreased the processing time of the scans. Because the data from each of the Kinects was pre-aligned to each other the processing time was typically about 5 minutes in KScan3D. About one out of every ten scans needed a bit of extra help though which usually meant an additional 5-15 minutes of effort to arrange the scans into a usable.

If we were just providing the raw figures this would have been it, but in order to give people something they could see and print right away we had to sink in a bit of extra effort. This worked out to about an average of 20 minutes per scan to produce a low poly version with a texture map that could be viewed online as well as a version with a base that could be printed via Shapeways.

Shawn and Alfred from FITC

So all in it averaged out to about 28 minutes per scan, including 2 minutes of scanning time and one in ten scans taking an additional 10 minutes of processing time. For 331 scans that nets out at 9,268 minutes. In other words 154.5 person hours or almost four person/weeks. Which has kept us busy.

If you add in some time for coordination, the build of the scanner and the site that displays the models and it’s likely total effort for for everything involved was closer to 6 weeks.

Why the hell did we do this you might ask? Because we could, because it was cool, and because no one else has done it. Like I said before, this is what we do.

Will we do something like this again? Undoubtedly.

Clayton Partridge's picture
BY: Clayton Partridge
Founding Partner

Clayton is a developer, and likes to make things out of other things.

comments powered by Disqus