3D I, Week 8, Part 2: 3D Scanning

This is my final blog report for the 3D I class. In this blog, I will talk about the process of facial scanning and reflect on what we’ve learned about face topology.

Facial scanning is the act of taking photos of a person and throwing them into a software to convert them into a 3d model. This requires a huge amount of photos, the more photos you get the more accurate the generation will be. The software we used in class to generate models was Autodesk 123D Catch.

Examples of photos that could be used for a 3D face scan.

The photos taken require incredible consistency for the program to read them right. The lighting needs to be low contrast and match between every photo, and if the person depicted in the photos moves a single muscle, the program will be confused when constructing the model. As such, when 3D scanning human faces, usually a multiple camera setup is used to take multiple pictures from different angles at the same millisecond.

An example setup for capturing a face from several angles at the same time.

In the classes of week 8, we went through how to model a face, and what the edgeflow should look like to make for a model that looks good when moving. The basic structure we learned is shown in the picture below. The most important parts of this structure are the circles formed around the eyes and the mouth.

We had pictures taken of our own faces on the Tuesday lesson of week 8, which we were to then 3d scan and work off of in the Wednesday lesson to practice topology of the human face. The photos for most people ended up too blurry and inconsistent to use, including mine, so I used a standard face scan provided during the lesson.

We worked in 3D Coat and went over the face scans provided to create lower poly models. 3D Scans often provide models with millions of triangles, so they have to me remade to be usable. I got as far as modeling out the circles around the eyes and mouth in my retopology.

This method of creating circles around these parts of the face is done with the natural deformations of the face in mind. Opening and stretching out the mouth on a model made with this method will stretch the skin in surrounding areas of the face in a more natural way than if you simply modeled out the face with a more consistent checkerboard-like topology.

3D scanning can be used on more than just human faces. It can used for virtually any existing object or creature in the world as long as you can get enough consistent photos from different angles for the program to construct a model. 3D scanning a rock or a building is generally easier than scanning a human, since these objects don’t move, but it still requires a great deal of fidelity. When taking photos for 3D scanning, generally the same rules apply as taking stock photos for texture usage. You want even, low contrast lighting, no strong shadows or highlights, no extreme depth of field, and absolutely no movement. Even with static objects, achieving all of these things can be difficult.

3D scanning, when executed well, gives you a high quality, realistic model of a chosen object. But with the amount of photos required, the remodeling work, and any eventual adjustment at any stage in the production, makes it time consuming. It’s an effective tool when you’re striving for photorealism and you want to capture natural details in your model that would take huge amounts of work to replicate when working from scratch.