Introduction: Steel Masks Based on a 3D Face Scan

About: A kinetic sculptor known as Fish. He is currently making a slow, terrifying transition from computer professional to full-time artist.

I've spent a lot of the last year experimenting with a process to create tessellated steel forms using 3D models and waterjet cutting/etching. (More details given here and here.) I'm going to be using it scaled up in some public art installations soon, but I first wanted to push it in the other direction. How finely detailed could I make things before it just stopped working?

Step 1: Modeling

First I needed a scan of my face. I did some searches for providers in the Seattle area, and the results weren't encouraging. They were all very high end services, the kind that don't even list their prices. They obviously didn't want to work with someone on a cheap little weird project. Then I discovered 3D Hubs which lets you find local people with printing or scanning equipment. Soon I was sitting in the food court of a mall while someone waved a Kinect-style scanner around my head for cash. Perfect!

The results were pretty good, too. As you can see, my pony tail confused it quite a bit, and the muttonchops are a bit weird. But the important parts, like the eye/nose topography, all came through quite well. I inflated the entire thing by 1 mm using the Blender solidify modifier, to allow for a bit of error.

The next step was to create masks based on the pointcloud data. After experimenting with various Blender modifiers, I ended up doing it manually. For each mask I would create a copy of the pointcloud, then enter edit mode and show the vertices. I selected all the ones that I wanted as a corner of the mask, then deleted the rest. Figuring out which ones to choose took some trial and error. You have to remember that a flat steel plane will be connecting each one, and you don't want that intersecting your face! But don't stress too much about it, because you can tweak them in the next step.

Once you have the points, connect them into a reasonable mesh. Now move the mesh back over the original face scan. This allows you to see where the face scan surface intersects the mask. Adjust the points until that doesn't happen. If you want a symmetrical design, like with the domino mask I made, you can design only half and then mirror it across the Z axis.

Step 2: Assembly

After finishing the 3D models in Blender, I unfolded them using Pepakura and sent them off to be waterjet cut and etched. Please see my other Instructables for more details on that process.

Unlike my previous experiments with this technique, I had no guide for assembly. There were plenty of unconstrained vertices where more than 3 faces met. I didn't have a 3D printed internal form to test the angles against. This is the step I knew could easily fail, particularly with so many fiddly little bits around the nose and lips on the full mask.

In the end it worked, but not always particularly well, and it took a lot of trial and error. The two smaller masks weren't so bad, and the gaps weren't too big on them. The full mask was not nearly so kind. The problem is that errors accumulate, so the joins just kept getting worse and worse. The last bits of the full mask were multiple centimeters out of alignment by the time they met up. I ended up having to make create modifications with an angle grinder to make it work.

If I were to do it again, I'd be a lot more careful when breaking the design into sections to make sure the joins all followed the sharpest concave bends. (The "valley fold", to use Pepakura's terminology.) The etched steel can only bend so far in that direction, so putting a section break there prevents a lot of complications. I'd also think seriously about spending the extra money to 3D print a version of the design, so I could get all the bends just right.

Step 3: Finishing

For the 2 smaller masks, I used the same finishing technique I've used in the past. I sandblasted them to remove the mill scale, then I carefully brushed gun bluing solution onto the faces, not letting it get down into the cracks. After scrubbing it off the next day, I buffed the surface and sprayed it with a clear lacquer.

I decided to do something different for the full mask, mostly because painting on the gun bluing solution is really tedious. I hit it with a propane torch, carefully heating it up into the 200-300 C range to develop tempering colors. You can't be very precise with a torch like that, so I just went for a random series of splotches in the various colors. It made me wish I had the set up to go for a uniform deep blue, so that might be a future project. This was followed again with the clear lacquer.

Unfortunately, while the steel was looking pretty good at this point, there were still some very distracting gaps where the pieces hadn't lined up perfectly. I'm never above cheating, so I glued black felt onto the backsides of each mask with E6000. This radically reduces how visible the gaps are. They are now solidly in the category of things that still deeply bug me but not one else really notices. So, yay?

Straps were sewn onto the smaller masks, and a shiny stainless steel rod was welded to the full mask so it could be held in place. (Since it covered my mouth, I didn't want it tied down!) Feeling fancy, I turned a nice brass handle for the rod on my lathe. The rounded notches were made by grinding a radius into the end of a 1/2" tool blank and plunging it in. Brass is soft enough to get away with doing that, plus a bit of sanding afterwards to smooth it out.

Step 4: Results

They aren't as good I was hoping, but actually a bit better than I expected. They have already proved useful at a masked ball (image courtesy of Andrew Ferguson), and will also do nicely for my company costume contest and Halloween party coming up. And, more importantly, they were entirely successful as an experiment.

I now know you can use this technique on tiny, fiddly details. If you don't care too much about gaps, you don't even have to use a reference when bending the angles. You can design the models based on 3D scans of existing objects and get a reasonably close fit. Thinking through the geometry is even more critical with this level of complexity, however, particularly on larger designs where subsections are being meeting each other from all directions.

Cheers, all. Time to get busy on bigger projects for a bit. Look for my next Instructable "Designing and Building Public Art" to drop (hopefully) in early December!

Halloween Costume Contest 2015

Participated in the
Halloween Costume Contest 2015