In this recipe, we used two different models: one for gender classification and one for age group classification. Note that in this recipe, in contrast with the others, we subtract per-pixel mean values from the source image, not per-channel values. You can actually visualize the mean values and see the average human face.
Here's the input image:
The following output is expected:
Gender: female with prob: 0.9362890720367432
Age group: (25, 32) with prob: 0.9811384081840515