Tuesday, 1 July 2014

MakeHuman – Tests with facial reconstruction and human evolution

Face of a reconstruction turning into an Australopithecus afarensis. Model originally created in MakeHuman and Blender later deformed with ShapeKeys. Animated GIF and effects generated with Imagemagick.

I remember in late 2011 I bought a classic book called Forensic Art Art & Illustration. After studying the process of facial reconstruction I decided to apply it on a skull acquired from a CT scan, and to my happiness, it all went successful.

The year 2012 was prolific in reconstructing faces, and by 2013 I developed a methodology for facial reconstruction based on pre-structured models, in partnership with Dr. Paulo Miamoto. We aimed at a more rapid completion of work and more than that, we also in thought about providing the beginner with the ability to make their own reconstructions without the need to study for years in order to develop the mastery of sculpting a 3D face digitally.

Face adapted from a photo background view. The goal was to only define the volumetric shape, without setting beard, hair and accessories. The terracotta face on the left is the standard model of MakeHuman.
Even with the development of a pre-configured template, the task of revealing faces from skulls was still difficult for those who were starting and it forced us to look for more reasonable alternatives. That's when MakeHuman appeared.

First of all it must be explained how our "assembly line" of faces works. Initially I get a skull, usually without much information. When I say "get a skull" in fact it is either a CT or a laser scan. Then I send the material to Dr. Miamoto, who observes it and estimates the sex, age and the ancestry of the individual, based on knowledge from forensic protocols.

Face adaptation in MakeHuman, using a background image as a reference.
For many months, Dr. Miamoto and I did hours and hours of virtual meetings, so called "hangouts", in which we exchanged information. He handed me knowledge inherent to Forensic Dentistry and Anthropology and I taught him how to use 3D computer graphics open software. Gradually he learned the processes, and started handing me the skull aligned and with the soft tissue markers placed (those little pegs on the surface of the skull that are estimates of the thickness of the skin, fat and muscle in the region).

Just the fact that he handed me this stage of work already done quickened the wprkflow considerably. Our challenge became another step... the basic modeling of the face. Even with all the knowledge gained in computer graphics, Dr. Miamoto saw himself struggling to handle the template we had developed, because even though it was preset, handling it demanded a some ability of the user, which could only be gained by many months of dedication.

The luck turned to our side at the moment I received the news of the new version of the free software MakeHuman. Not that I had not used it before, in fact I knew it fairly well, but I didn't find it very suitable in previous versions, due to some incompatibilities on importation of human models in Blender. However, as I tested the scripts provided by the developers I had a pleasant surprise as I realized that everything worked perfectly.
For those unfamiliar, we can sey long story short that MakeHuman is a "human factory" tool. When we open the software we are presented to a standard model which may be adapted to any gender, age or ancestry just by manipulating a series of attributes intuitively organized into its interface. Everything visually and in real time.

In theory we can create any type of human being, and the most interesting is that given its ease of use, anyone can operate it in a few minutes.

The idea that came to mind was that, as Dr. Miamoto had dominated the first part of the process he could also set up the model based on the anthropological profile assessed from the skull, and send me the file so that I would use it to adapt the shape defined by skull itself, the markers in the nasal tissue and other projections.
The first test I did was with a skull that I had reconstructed two years ago. I sent the skull, he observed it and assessed sex, age and ancestry. Then he set up a model in MakeHuman based on the data collected and clues provided by the skull. Once finished he sent it to me via email.

Back in 2012 when modeled it from scratch, that face took about 8 hours to be done. By adapting the face according to the data received from Dr. Miamoto it did not take me more than 20 minutes to finish the volumetric face! I was amazed after I imported the mesh that I had done in 2012 and verified that the two were very compatible with each other.
The test was successful, and the more interesting... it was not only the head, but an already articulated whole body, ready to be animated in case it was necessary!

Skull aligned and profile sketched by Dr. Miamoto
Above, an example of soft tissue pegs being positioned over a skull and then the outline of the profile, made by Dr. Miamoto.

Adaptation process of the 3D face set in MakeHuman and then imported into Blender, done by Dr. Miamoto.
After setting up the face of the individual based on the anthropological assessments, he imported the model of MakeHuman into Blender and gradually adapted the facial tissue according to the projections of the profile.
Final stage of the adaptation of the face profile using as reference the layout of the profile.
Note that Blender and other programs are running under Linux. According to him, it made ​​the job much easier and practical.
Below are some words of Dr. Miamoto himself about MakeHuman and the proposed new methodology for digital facial reconstruction addressed in this post:
Fourth class of the graduate course in Forensic Dentistry at the Faculty of Dentistry of Ribeirão Preto (FORP-USP). Classes on forensic facial reconstruction with open software.
Dr. Miamoto explaining how photography-based 3D scanning works.
Below are some words of Dr. Miamoto himself about MakeHuman and the proposed new methodology for digital facial reconstruction addressed in this post:

"In Brazil, increasingly, the postgraduate courses in Forensic Dentistry include in their curriculi contents on forensic facial reconstruction. Generally, students are initiated in the manual technique and soon feel motivated to deepen their knowledge. It's an almost instinctive curiosity about the digital techniques, often with romaticized views (influenced by TV series), in which it is believed that there is an unique software that does all the work and everything is automatic. There are also intimidated views, as if the digital world is an extremely hostile terrain for "outsiders" such as dentists. The partnership that Cicero and I are working, always with the broad support of many partners in Brazil and abroad, helps make teaching beginners the digital techniques reality. The protocol we developed is constantly improving, and undoubtedly MakeHuman constitutes a keystone, which takes the weight of digitally modeling a very coherent face from a few polygons off the back of beginners (and also the experienced). In other words, although artistic skills are and always will be desirable, it is possible to start to work in 3D environment even if one is not an extremely skilled forensic artist. In addition, MakeHuman's intuitive interface combined with its excellent software development also allows for a real demonstration of the key features of each ancestry, differences between sizes and shapes of faces and cranial vaults, development aging of the human body, as well as other nuances of the human variation that are very useful to the teaching of Forensic Dentistry and Forensic Anthropology. By mastering some functions and basic concepts of Blender, the avatar that is imported from MakeHuman can be adapted to the skull with relative ease, while maintaining the characteristics of sex, age and ancestry, providing the student with the crystallization of a technique regarded as very complex or difficult, namely, computerized forensic facial reconstruction. The satisfaction and pride that a student feels when actually modeling a face on the computer is a spark that we believe to be a possible trigger for changes in the panorama of Brazilian forensic science, which certainly has in the motivation and training of its own human resources one of the cornerstones for its growth and consolidation. Thanks to the work of the open source software comunity, new solutions to old problems can be glimpsed, with the advantage of the wide accessibility to tools as well as their low or even nonexistent cost. Long live the open software applied to Forensic Sciences!”

Paulo Eduardo Miamoto Dias, DDS, MDS, PhD
CROSP 91.834

To make it an even more interesting task and prove once and for all that the job gets easier, I decided to test the deformation of the model we reconstructed from MakeHuman in a hominid.

I used an Australopithecus afarensis reconstruction, performed a few months ago for shows in Brazil and Italy, as a reference.

Using Blender's shapekey, which allows one to deform an object preserving the information from the original object, it is possible to track the "morph" of the face.

Technically we can convert the model into many other mammals, like a saber-tooth tiger, a mouse etc., not only other hominids.

If we can make such a different adaptation like this, why would the mild conversion of a human face into another be a problem?

The most interesting is that as we deal with open software with shareable content, there is a great possibility of establishing a partnership with MakeHuman's development team to collaborate on this very interesting project, which makes it accessible to the general public this art that is seen as extremely complex, which is the facial modeling.

Breathing the atmosphere of the World Cup, I would say that the kickoff was given, that contacts were made and that now everything depends on a little time and effort to get things to start happening.

See you next post!

Obs.: Thanks to Dr. Miamoto for the translation to English, from the original post in Portuguese.
BlogItalia - La directory italiana dei blog Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.