PhotoFit

[Katy Burgess]

Dear FaceGen team

I am looking for a bit more information on how the PhotoFit aspect of your software fits an image to the generated face. That is, how does FaceGen project the shape information onto the generated face. Does it project onto a generic shape template that is derived from the original laser scans?

Thanks in advance for your advice

Katy

[FaceGen Guru]

Dear Katy,

FaceGen matches photos by "fitting" our statistical model of shape and color, along with lighting and camera models, to make the best possible match to the photograph.

So no, we are not using a single template shape during this process but in fact a statistical model of shape variation encompassing many shape parameters.

Although this process is slower than other methods, it yields better results. Further, it allows for cool post-editing; for example of age, race and gender.

Best wishes,

[Katy Burgess]

Dear FaceGen,

Thanks for your speedy response. Just to follow up, does this mean that if the lighting of my input image was different but the image was the same the modeler would yield different results?

Kind regards

Katy

[FaceGen Guru]

Hi Katy,

It doesn't mean that in theory, because it solves for the lighting and face color as separate variables, then uses just the face color.

In practice however, out statistical models don't perfectly match reality, so yes, the lighting in the face photo will make some difference on the face color.

Best wishes,