Digital Human Experiments…

Took me a while to get this update together! This time, I’d like to share an experiment in single camera photogrammetry. I was very inspired by some of James Busby’s recent work on Sketchfab, so much so that I went back to try scanning with another human subject. My Dad dropped by the house one evening and I asked him to sit for me outside, just as the light was fading. I was fairly sure the effort would be wasted due to the failing light, but photogrammetry has surprised me before, so I carried on. I shot about 45 photos in raw, handheld on my phone, a Samsung Galaxy S8. From there, I processed the raw files in a free version of DxO, removing vignetting, chromatic aberration, shadows and highlights where possible, but keeping distortion. Here is the mesh output on normal detail in Reality Capture.

Results out of the box showed some promise. Overall form and volume were there with some detailed areas around the eyes quite well defined. Lots of noise too though, a chunk missing under the chin and nothing at all from the ears back. All in, there was a count of about 45 million triangles.  Generating the albedo texture helped the visuals a lot (see below) but I knew there’d still be cleanup in Mudbox, which has good hole patching for geometry. The relax, smooth and scrape brushes also work well for reducing noise in rough areas.
I used Instant Meshes to retopologise the high poly mesh, it can be a very handy tool to put out an all quad mesh to work with in real-time, here’s how the model looked, I also replaced the eyes with proper spherical eyeballs.
I laid out UVs on the low poly mesh in 3DSmax and baked albedo, thickness and normal maps from there, over then to Knald to generate high frequency detail normals, AO and cavity maps. Substance Painter and Photoshop were used to paint out shadows and highlights and fill in gaps in the textures, and hand paint specular and glossiness maps, to control which parts of the skin would look oily. I used Marmoset Toolbag 3 for look development, check out some of the textures below.
I still needed something to cover the large hole in the back of the mesh, I scanned a fur lined hat and tried to make it fit the head but I wasn’t happy with the results, I had to deform it too much to make it fit (image rendered in Toolbag 3)
I also tried a metal morion helmet I had already made, but it wasn’t great either and hid too much of the model underneath (image rendered in Toolbag 3)
So I ended up modelling a tight fitting wooly hat over the top of the head and textured it traditionally, and I was a lot happier with this. Here is the head model screen grabbed from Sketchfab:
Another (Sketchfab) image including the hat:
And finally the model in real-time, you can orbit by left-clicking and dragging or move the lights by holding alt and left-clicking and dragging. There is a very simple eye movement animation and a roll of the jaw, I’d love to try some more with facial rigging after this but it’s an area I know very little about. Lots learned from this one!

Bonus: Blooper with displacement on the wooly hat 🙂
SmartSelectGIF_2017-11-22-23-06-36

Advertisements

~ by petemcnally on November 23, 2017.

4 Responses to “Digital Human Experiments…”

  1. Great job Pete.

    ________________________________

  2. Thanks Joe!

  3. Very nice job.
    You might be interested in see these face model:
    https://skfb.ly/XHWt

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

 
%d bloggers like this: