Systems and methods for artificial facial image generation conditioned on demographic information
A method, computer program product, and computer system for analyzing, by a computing device, a plurality of facial images to determine a plurality of demographic labels associated with each facial image of the plurality of facial images. A model may be trained based upon, at least in part, the plurality of demographic labels associated with each facial image of the plurality of facial images. An input of at least a portion of the plurality of demographic labels may be received. An artificially generated facial image may be provided for display that is generated based upon, at least in part, the model and the input.
1. A computer-implemented method comprising:
analyzing, by a computing device, a plurality of facial images to determine a plurality of demographic labels associated with each facial image of the plurality of facial images;
training a model based upon, at least in part, the plurality of demographic labels associated with each facial image of the plurality of facial images, wherein the model is a generative adversarial network (GAN) with a generator and a discriminator, wherein the generator includes a label embedding layer that processes a one-hot encoded vector for each of the demographic labels, wherein the discriminator includes an affine layer that processes a tensor for each of the demographic labels;
receiving an input of at least a portion of the plurality of demographic labels; and
providing an artificially generated facial image for display that is generated based upon, at least in part, the model and the input.
2. The computer-implemented method of claim 1 wherein the demographic labels include at least one of a gender, an age, and an ethnicity.
3. The computer-implemented method of claim 2 wherein the age is based upon, at least in part, a weighted blending of age labels of the plurality of demographic labels.
4. The computer-implemented method of claim 2 wherein the ethnicity is based upon, at least in part, a weighted blending of ethnicity labels of the plurality of demographic labels.
5. The computer-implemented method of claim 1 wherein the plurality of facial images are automatically selected for analysis from an internet.
6. The computer-implemented method of claim 1 further comprising determining whether a candidate face image is acceptable for use as one of the plurality of facial images to train the model based upon, at least in part, at least one of whether the candidate face image is a real human face and one or more attributes of the candidate face image.
7. The computer-implemented method of claim 1 wherein the model is trained to automatically assign the demographic labels to the plurality of facial images.
8. A computer program product residing on a non-transitory computer readable storage medium having a plurality of instructions stored thereon which, when executed across one or more processors, causes at least a portion of the one or more processors to perform operations comprising:
analyzing a plurality of facial images to determine a plurality of demographic labels associated with each facial image of the plurality of facial images;
training a model based upon, at least in part, the plurality of demographic labels associated with each facial image of the plurality of facial images, wherein the model is a generative adversarial network (GAN) with a generator and a discriminator, wherein the generator includes a label embedding layer that processes a one-hot encoded vector for each of the demographic labels, wherein the discriminator includes an affine layer that processes a tensor for each of the demographic labels;
receiving an input of at least a portion of the plurality of demographic labels; and
providing an artificially generated facial image for display that is generated based upon, at least in part, the model and the input.
9. The computer program product of claim 8 wherein the demographic labels include at least one of a gender, an age, and an ethnicity.
10. The computer program product of claim 9 wherein the age is based upon, at least in part, a weighted blending of age labels of the plurality of demographic labels.
11. The computer program product of claim 9 wherein the ethnicity is based upon, at least in part, a weighted blending of ethnicity labels of the plurality of demographic labels.
12. The computer program product of claim 8 wherein the plurality of facial images are automatically selected for analysis from an internet.
13. The computer program product of claim 8 wherein the operations further comprise determining whether a candidate face image is acceptable for use as one of the plurality of facial images to train the model based upon, at least in part, at least one of whether the candidate face image is a real human face and one or more attributes of the candidate face image.
14. The computer program product of claim 8 wherein the model is trained to automatically assign the demographic labels to the plurality of facial images.
15. A computing system including one or more processors and one or more memories configured to perform operations comprising:
analyzing a plurality of facial images to determine a plurality of demographic labels associated with each facial image of the plurality of facial images;
training a model based upon, at least in part, the plurality of demographic labels associated with each facial image of the plurality of facial images, wherein the model is a generative adversarial network (GAN) with a generator and a discriminator, wherein the generator includes a label embedding layer that processes a one-hot encoded vector for each of the demographic labels, wherein the discriminator includes an affine layer that processes a tensor for each of the demographic labels;
receiving an input of at least a portion of the plurality of demographic labels; and
providing an artificially generated facial image for display that is generated based upon, at least in part, the model and the input.
16. The computing system of claim 15 wherein the demographic labels include at least one of a gender, an age, and an ethnicity.
17. The computing system of claim 16 wherein the age is based upon, at least in part, a weighted blending of age labels of the plurality of demographic labels.
18. The computing system of claim 16 wherein the ethnicity is based upon, at least in part, a weighted blending of ethnicity labels of the plurality of demographic labels.
19. The computing system of claim 15 wherein the plurality of facial images are automatically selected for analysis from an internet.
20. The computing system of claim 15 wherein the operations further comprise determining whether a candidate face image is acceptable for use as one of the plurality of facial images to train the model based upon, at least in part, at least one of whether the candidate face image is a real human face and one or more attributes of the candidate face image.
21. The computing system of claim 15 wherein the model is trained to automatically assign the demographic labels to the plurality of facial images.