PHOTOGRAPHY Revisiting Virtual Studio Idea With VIZAGO

I stumbled into this Vizago 3D Head Reconstruction website today that gives a demo that allow user to upload a single portrait photo (frontal shot) and then it tries to convert it into 3D mesh object.

Full of curiosity, I quickly pickup a random photo from Google and I grabbed a portrait from here:

VIZAGO 3D HEAD RECONSTRUCTION
Upload a single portrait photo into Vizago website.

Get 3D head mesh within a couple of minutes or a day (depends on the demands, I supposed)
It comes to my surprise that the result is actually quite nice. And you can preview the result in 3D on the web directly. You can rotate around the head mesh on browser as well.

In the past, I remember there are already few casual/hobby applications that does something like this, but the result was not great. Now you can do it really quickly on a website and at this demo stage, you can download the resulting OBJ complete with texture and UV. To download, you need promo code, just ask Vizago via email.

IMPORT TO BLENDER
When you bring this model into Blender, you may want to scale down the mesh x 0.0001 (or x 0.00001). The mesh was a bit too big to fit the scene.
  • As you import the OBJ, just tap S and then type 0.0001.
  • Scale down even more if still does not fit.
  • You then want to center the pivot, simply Right Click on the mesh (to select it) and CTRL+SHIFT+C and select Origin to Geometry and the pivot will now centered to the mesh geometry.
  • To freeze the Scaling, CTRL-A and Apply Scale.
This is fun for you all Blenderheads to try. Maybe create a 3D head of yourself, friends, or family and then do 3D printing, using service like Shapeways.

Or if you just like to play around with the mesh inside Blender, probably creating some facial animation using Blender Shapekeys, you can also do that. Perhaps create all kind of hair setup.

PAT DAVID VIRTUAL STUDIO FOR BLENDER
Remember my previous post on using Blender as Virtual Photography Studio? The one originally created by Patrick David?

Time to use the setup:
3D Head Model based on a single photo using the Vizago 3D Head Reconstruction

Below are some results:

Colorful lighting.

Killer mood.

Sensual mood like in a night club or inside Virgin Airplane.

Before Sunset mood.
Environment Lighting (image based lighting)

Just like when I first time use this setup, I thought this was really brilliant setup by Pat. Super useful to test all kind of lighting situations.


Yeah, unfortunately most of the models will be bald without hair. So, probably time to sculpt some hair!

Anyways, now that we have a web service like Vizago to create 3D head model automatically, you can test it on different kind of heads. Maybe even create 3D portrait (stereoscopic)?

Try all kind of head result using Martin‑Schoellers photography, just for educational purpose. See which one actually works and which one does not work.



So, maybe you can guess who are the female actress/model below:
Angelina Jolie has similar look to Miranda Kerr and Emily Browning, who would have thought. Jennifer Lopez actually has very distinct look.
Natalie Portman looks a little like Monica Belucci.
You can do interesting study of head facial using this process:
  • My quick guess is that they have template base head (one for male and one for female) that is morphed based on points you specified during the upload and reconstruction of the mesh.
  • Face features that are morphed: eyes, nose, lips and the main shape of the head (more round, more triangle). Lips and eyes are taken care mostly by textures. 
  • Nose is actually important in getting likeness. Probably they can improve if the user provides the NOSE PROFILE shape.
  • What interesting is how Vizago tries to guess the "profile" of the head.
  • The ears stay the same, probably just the position is different.

If you import some of the head mesh into Blender, noticed that the Vertex number of the mesh is the same. You can do "morphing" from one head to another. Shape Keys time!

UPDATE: The vertex number is apparently different. So you cannot quickly do Join as Shape Keys. But I think there is a way around this. Maybe Shrink Wrap Deformer and Vertex Weight Blending will do the job so that you can blend from one shape into another.

So far I only tried this using random photos that I found using Google Images, or Celebrity faces, which is really simple. Celebrity and model faces tend to be the most basic and simple. For faces that are more exotic, like islanders, probably the result will be more interesting. See if that works or not.


Doing this auto 3D head reconstruction process reminds me my previous experience on seeing this Chinese street artist doing small head sculpting from a real life model in realtime for about 30 minutes per head in Las Vegas. I thought that was really an amazing skill. Especially to get likeness to the persons face.

Eventually I think there will be lots of digital artist doing this manually in real time using Blender/Mudbox/ZBrush and print it out under 30 minutes. Right?

The applications like Vizago or perhaps the Autodesk 123DCatch that allows anyone to quickly take photos and create 3D mesh are amazing. You can use this as starting point reference, or use it right away.

Decide whether creating virtual avatar manually or automatically is more preferable? Here is another example from company in Korea "3D Avatar Generation": (probably for game figure)




Wonder if anyone like to use own figure for game or the better self?

Please show me your creations, it would be nice to see what you created with this!

Related Posts by Categories

0 komentar:

Posting Komentar