Optimizing PhotoFit: camera settings and point placement

[Adrian De La Mora]

Hi,

I want to figure out how to optimize the result of the PhotoFit feature before I can commit any resources into purchasing FaceGen software.

I am developing an animation pipeline around DAZ Studio and Motionbuilder, and our requirements are such that we need to build characters quickly, accurately and with consistency.  I have started playing with the demo versions of the Modeler and Exporter programs, and have come up with a few questions not clearly documented:

Pre-PhotoFit:

What are the best camera settings for getting the most accurate models?

- Wide-angle vs zoom? Is 50mm the best as it is the closes to the natural eye?

What composition should I have for the front/side photos? 

Changing the center point also changes the perspective of the edge features (chin, nose, cheeks, etc) so I assume the program wants to see photos in a certain way, otherwise, it will build features differently.

- How should the camera and subject be arranged? (camera angle, subject posing, etc.)

- Should the front photo be centered on the eyes/nose/mouth?

- Should the side photo be centered on the ears? (does PhotoFit even do anything with the ears?)

- Do I really need to provide both left/right profile photos? or can I simply use one and have it create a symmetrical face (I know, it will look kind of fake if it's symmetrical)

What lighting setup will result in the best texture files?

- Should the lighting be the same in both photos (rotate camera around seated subject)? Or can the subject be rotated in place (causing some light shifting)?

- Is there a particular background color that will work best? (we have a green-screen wall)

In-PhotoFit:

Where EXACTLY should the markers be placed on the images?

- I can see from the instructions that I am marking certain anatomical parts, but which are they specifically?

- Do some of the markers have to match up with markers on the other perspectives? is there a way to get a reference image of my previous marker placement to help me match them?

- Do I place the markers on the feature lines, or just outside/inside? what effect will this have on the final model/textures?

Post-PhotoFit

er.. no questions yet on this... winning!

I think you get the idea of what I am aiming at: I want to understand how each little action/setting results in the final model/texture.

Thanks!

[FaceGen Guru]

Hello Adrian,

For best results with the photofit you should be 5 to 10 feet from
your subject, so pick your lens appropriately.

You do not need to center the faces perfectly, but the region that you
should center is the 'face region'; from where the chin meets the neck
to half-way up the forehead, and in front of the ears.

If you use just one profile photo Modeller will use the details from
that photo for both sides of the face so there's no need to use 2
profile photos unless you have a specific scar or something that you
want to appear only on the one side.

Lighting is important. For best results use a consumer-style point
flash on the camera for all shots.

Background is not as important, although a dark background may help
with the outer contours.

For more details about taking the photo:

http://www.facegen.com/modeller35_help.htm#photofit

The marker point positions do not have to be exact, since FaceGen uses
them mainly as a starting point and has tolerance on their
positioning.

The points on the different views do not correspond.

For the frontal points: center of the pupils, the outermost points on
the nose flaps, corners of mouth, outermost points of the cheekbones,
occluding contour of the jaw at the same height as the mouth (or what
would be the same height if the face is tilted up/down).

For the profile points: sellion, nose tip, half way in-between on nose
bridge, outer corner of eye, where nose meets philtrum, forward-most
point of upper lip, forward-most point of lower lip, forward-most
point of chin, part-way (roughly) between chin and neck on lower chin
contour.

Best wishes,