Systems and methods for providing an image classifier

1. A computer-implemented method for image classification, the method comprising:

generating, from a set of training images, a non-rectangular foreground silhouette representing a relevant portion of an image window;

generating, from the silhouette, a non-rectangular bitmap identifying fewer than all cells within a rectangular detection window;

applying the rectangular detection window to a portion of the image having a cell width and a cell height of the rectangular detection window, wherein applying the detection window includes identifying cells within the image portion matching the identified cells of the non-rectangular bitmap;

for an identified cell within the image portion, analyzing orientations of pixels within the cell;

identifying an overlapping cell block having a plurality of contiguous identified cells from the image portion;

generating a descriptor for the cell block, wherein a contents of the descriptor is normalized to contents of another descriptor; and

evaluating the generated descriptor against the set of training images to determine whether to identify the image portion as including a class of object associated with the set of training images.

2. The method of claim 1 , further including:

re-scaling the image to a second cell width and cell height; and

applying the rectangular detection window to a portion of the re-scaled image having the cell width and cell height.

3. The method of claim 1 , further including:

averaging the training images; and

storing the averaged training images as the non-rectangular foreground silhouette.

4. The method of claim 1 , further including:

determining intensity values of pixels for the training images;

adding the intensity values of the pixels for the training images to generate resulting values;

dividing the resulting values by a number of the training images; and

storing the divided resulting values as the non-rectangular foreground silhouette.

5. The method of claim 1 , further including:

calculating gradients, the gradients having magnitudes and corresponding orientations;

associating the orientations to respective ones of a plurality of bins; and

adding the magnitudes to the respective ones of the bins associated with corresponding ones of the orientations.

6. The method of claim 1 , further including determining whether the class of object is one of:

a person;

a face;

a non-rectangular geometric shape; or

a non-geometric shape.

7. The method of claim 1 , wherein the description includes a histogram for the cell block, wherein the histogram is normalized to a histogram of the another descriptor.

8. A computing device for image classification comprising:

memory including:

a non-rectangular foreground silhouette generated based on a set of training images, and

a non-rectangular bitmap that identifies a subset of cells of a rectangular detection window, the rectangular detection window having a cell width and cell height;

a hardware accelerator to:

apply the rectangular detection window to a portion of the image having a cell width and a cell height of the rectangular detection window, wherein applying the detection window includes identifying cells within the image portion matching the identified cells of the non-rectangular bitmap;

for an identified cell within the image portion, analyze orientations of pixels within the cell;

identify an overlapping cell block having a plurality of contiguous identified cells from the image portion;

generate a descriptor for the cell block, wherein a contents of the descriptor is normalized to contents of another descriptor; and

evaluate the generated descriptor against the set of training images to determine whether to identify the image portion as including a class of object associated with the set of training images.

9. The computing device of claim 8 , wherein the hardware accelerator is to:

re-scale the image to a second cell width and cell height; and

apply the rectangular detection window to a portion of the re-scaled image having the cell width and cell height.

10. The computing device of claim 8 , wherein the hardware accelerator is to:

average the training images; and

store the averaged training images as the non-rectangular foreground silhouette.

11. The computing device of claim 8 , wherein the hardware accelerator is to:

determine intensity values of pixels for the training images;

add the intensity values of the pixels for the training images to generate resulting values;

divide the resulting values by a number of the training images; and

store the divided resulting values as the non-rectangular foreground silhouette.

12. The computing device of claim 8 , wherein the hardware accelerator is to:

calculate gradients, the gradients having magnitudes and corresponding orientations;

associate the orientations to respective ones of a plurality of bins; and

add the magnitudes to the respective ones of the bins associated with corresponding ones of the orientations.

13. The computing device of claim 8 , wherein the hardware accelerator is to determine whether the class of object is one of:

a person;

a face;

a non-rectangular geometric shape; or

a non-geometric shape.

14. The computing device of claim 8 , wherein determining the cell block includes determining a plurality of overlapping cell blocks having a plurality of contiguous identified cells.

15. A non-transitory computer readable medium comprising instructions that, when executed, cause a computing device to at least:

generate, from a set of training images, a non-rectangular foreground silhouette representing a relevant portion of an image window;

generate, from the silhouette, a non-rectangular bitmap identifying fewer than all cells within a rectangular detection window;

apply the rectangular detection window to a portion of the image having a cell width and a cell height of the rectangular detection window, wherein applying the detection window includes identifying cells within the image portion matching the identified cells of the non-rectangular bitmap;

for an identified cell within the image portion, analyze orientations of pixels within the cell;

identify an overlapping cell block having a plurality of contiguous identified cells from the image portion;

generate a descriptor for the cell block, wherein a contents of the descriptor is normalized to contents of another descriptor; and

evaluate the generated descriptor against the set of training images to determine whether to identify the image portion as including a class of object associated with the set of training images.

16. The non-transitory computer readable medium of claim 15 , wherein the instructions, when executed, cause the computing device to:

re-scale the image to a second cell width and cell height; and

apply the rectangular detection window to a portion of the re-scaled image having the cell width and cell height.

17. The non-transitory computer readable medium of claim 15 , wherein the instructions, when executed, cause the computing device to:

average the training images; and

store the averaged training images as the non-rectangular foreground silhouette.

18. The non-transitory computer readable medium of claim 15 , wherein the instructions, when executed, cause the computing device to:

determine intensity values of pixels for the training images;

add the intensity values of the pixels for the training images to generate resulting values;

divide the resulting values by a number of the training images; and

store the divided resulting values as the non-rectangular foreground silhouette.

19. The non-transitory computer readable medium of claim 15 , wherein the instructions, when executed, cause the computing device to:

calculate gradients, the gradients having magnitudes and corresponding orientations;

associate the orientations to respective ones of a plurality of bins; and

add the magnitudes to the respective ones of the bins associated with corresponding ones of the orientations.

20. The non-transitory computer readable medium of claim 15 , wherein the instructions, when executed, cause the computing device to determine whether the class of object is one of:

a person;

a face;

a non-rectangular geometric shape; or

a non-geometric shape.