Style transfer lies outside my specialization, so take my comment as speculation informed by intuition from variously related works that may not correctly carry over to this one.
If you look at the nature of the transformations achieved in this paper, you might note that they are changes in, well, style. That is, the presentations of the objects in an image are represented using a different _style_, but they remain the same object.
As an example, take a look at the horse/zebra transformation; the horse obtains a zebra's stripes, but it _structurally_ still looks like a horse. That is to say, a zebra and a horse have identifiably different bodily proportions, and the horse's bodily proportions are not changed by the style transfer. Similarly, the trees in the summer -> winter transformation do not have the slightly saggy branches that they would due to the weight of snow on them.
With that in mind, I would be surprised if the approach taken in this work, taken as-is, would be able to change the gender or ethnicity of a person in a photo. There are structural differences between men's and women's faces, and similarly between races. I would imagine that an attempt of a race transfer, as I suppose you would call it, would largely amount to changing skin tone.
The age progression application, though, might work under the limitations I have speculated, at least for aging photos of a person that is already more or less mature. A person's facial structure does not significantly change once adulthood is reached, so simply transferring stylistic features might add wrinkles and other age-related changes in a realistic way.
Again, speculation. And even if the speculation is correct, that is not to say that some modifications to the approach would not be able to lift the constraints I made up.
Edit: reading through the rest of the paper[0] behind this work, looks like I might not be far off:
>Although our method can achieve compelling results in
many cases, the results are far from uniformly positive. Several
typical failure cases are shown in Figure 17. On translation
tasks that involve color and texture changes, like many
of those reported above, the method often succeeds. We
have also explored tasks that require geometric changes,
with little success.
The quote is from Section 6 ("Limitations and Discussion"), and example "limitations" are given in Figure 17.
If you look at the nature of the transformations achieved in this paper, you might note that they are changes in, well, style. That is, the presentations of the objects in an image are represented using a different _style_, but they remain the same object.
As an example, take a look at the horse/zebra transformation; the horse obtains a zebra's stripes, but it _structurally_ still looks like a horse. That is to say, a zebra and a horse have identifiably different bodily proportions, and the horse's bodily proportions are not changed by the style transfer. Similarly, the trees in the summer -> winter transformation do not have the slightly saggy branches that they would due to the weight of snow on them.
With that in mind, I would be surprised if the approach taken in this work, taken as-is, would be able to change the gender or ethnicity of a person in a photo. There are structural differences between men's and women's faces, and similarly between races. I would imagine that an attempt of a race transfer, as I suppose you would call it, would largely amount to changing skin tone.
The age progression application, though, might work under the limitations I have speculated, at least for aging photos of a person that is already more or less mature. A person's facial structure does not significantly change once adulthood is reached, so simply transferring stylistic features might add wrinkles and other age-related changes in a realistic way.
Again, speculation. And even if the speculation is correct, that is not to say that some modifications to the approach would not be able to lift the constraints I made up.
Edit: reading through the rest of the paper[0] behind this work, looks like I might not be far off:
>Although our method can achieve compelling results in many cases, the results are far from uniformly positive. Several typical failure cases are shown in Figure 17. On translation tasks that involve color and texture changes, like many of those reported above, the method often succeeds. We have also explored tasks that require geometric changes, with little success.
The quote is from Section 6 ("Limitations and Discussion"), and example "limitations" are given in Figure 17.
[0] https://arxiv.org/pdf/1703.10593.pdf