Landmark-free face attribute prediction

1. A computer-implemented method for landmark-free facial attribute recognition in digital images, the method being executed by one or more processors and comprising:

receiving an input image comprising at least one face;

processing the input image through a global transformation network to provide a set of global transformation parameters;

applying the set of global transformation parameters to the input image to provide a globally transformed image;

processing the globally transformed image through a global representation learning network to provide a set of global features;

processing the set of global features through a part localization network to provide a set of part localization parameters;

applying the set of part localization parameters to the globally transformed image to provide a locally transformed image;

processing the locally transformed image through a part representation learning network to provide a set of local features; and

outputting a label representing at least one attribute depicted in the input image based on fusing at least one global feature from the set of global features, and at least one local feature from the set of local features.

2. The method of claim 1 , wherein the set of global transformation parameters are tailored for the input image for attribute representation learning, and a transformation is provided based on the set of global transformation parameters that establishes a mapping between the input image and the globally transformed image.

3. The method of claim 1 , wherein the global representation learning network maps the globally transformed image from raw pixel space to a feature space.

4. The method of claim 1 , wherein the global transformation network, and the global representation learning network are trained together end-to-end to minimize an attribute predication loss.

5. The method of claim 1 , wherein the set of part localization parameters are applied to position a focus window to a relevant part on the at least one face through learned scaling and translating transformations.

6. The method of claim 1 , wherein the set of part localization parameters is specific to an attribute of a plurality of attributes.

7. The method of claim 1 , wherein multiple attributes of a plurality of attributes share the set of part localization parameters.

8. A non-transitory computer-readable storage medium coupled to one or more processors and having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to perform operations for landmark-free facial attribute recognition in digital images, the operations comprising:

receiving an input image comprising at least one face;

processing the input image through a global transformation network to provide a set of global transformation parameters;

applying the set of global transformation parameters to the input image to provide a globally transformed image;

processing the globally transformed image through a global representation learning network to provide a set of global features;

processing the set of global features through a part localization network to provide a set of part localization parameters;

applying the set of part localization parameters to the globally transformed image to provide a locally transformed image;

processing the locally transformed image through a part representation learning network to provide a set of local features; and

outputting a label representing at least one attribute depicted in the input image based on fusing at least one global feature from the set of global features, and at least one local feature from the set of local features.

9. The computer-readable storage medium of claim 8 , wherein the set of global transformation parameters are tailored for the input image for attribute representation learning, and a transformation is provided based on the set of global transformation parameters that establishes a mapping between the input image and the globally transformed image.

10. The computer-readable storage medium of claim 8 , wherein the global representation learning network maps the globally transformed image from raw pixel space to a feature space.

11. The computer-readable storage medium of claim 8 , wherein the global transformation network, and the global representation learning network are trained together end-to-end to minimize an attribute predication loss.

12. The computer-readable storage medium of claim 8 , wherein the set of part localization parameters are applied to position a focus window to a relevant part on the at least one face through learned scaling and translating transformations.

13. The computer-readable storage medium of claim 8 , wherein the set of part localization parameters is specific to an attribute of a plurality of attributes.

14. The computer-readable storage medium of claim 8 , wherein multiple attributes of a plurality of attributes share the set of part localization parameters.

15. A system, comprising:

a computing device; and

a computer-readable storage device coupled to the computing device and having instructions stored thereon which, when executed by the computing device, cause the computing device to perform operations for landmark-free facial attribute recognition in digital images, the operations comprising:

processing the input image through a global transformation network to provide a set of global transformation parameters;

applying the set of global transformation parameters to the input image to provide a globally transformed image;

processing the globally transformed image through a global representation learning network to provide a set of global features;

processing the set of global features through a part localization network to provide a set of part localization parameters;

applying the set of part localization parameters to the globally transformed image to provide a locally transformed image;

processing the locally transformed image through a part representation learning network to provide a set of local features; and

outputting a label representing at least one attribute depicted in the input image based on fusing at least one global feature from the set of global features, and at least one local feature from the set of local features.

16. The system of claim 15 , wherein the set of global transformation parameters are tailored for the input image for attribute representation learning, and a transformation is provided based on the set of global transformation parameters that establishes a mapping between the input image and the globally transformed image.

17. The system of claim 15 , wherein the global representation learning network maps the globally transformed image from raw pixel space to a feature space.

18. The system of claim 15 , wherein the global transformation network, and the global representation learning network are trained together end-to-end to minimize an attribute predication loss.

19. The system of claim 15 , wherein the set of part localization parameters are applied to position a focus window to a relevant part on the at least one face through learned scaling and translating transformations.

20. The system of claim 15 , wherein the set of part localization parameters is specific to an attribute of a plurality of attributes.