Method of analyzing objects in images recorded by a camera of a head mounted device

1. A method of analyzing objects in images recorded by a camera of a head mounted device, the method comprising:

performing eye tracking while recording the images;

determining a region of interest of an image based upon the eye tracking;

generating a bounding box based upon the region of interest;

cropping an image based upon the bounding box to generate a cropped image;

performing a fine cropping of the cropped image to generate an updated region of interest; and

detecting an object in the cropped image after the fine cropping.

2. The method of claim 1 , wherein detecting an object in the cropped image comprises performing deep learning.

3. The method of claim 2 , further comprising generating an object detection confidence value associated with the object in the image, and using the object detection confidence value to determine a region of interest of a later image.

4. The method of claim 1 , wherein performing a fine cropping of the cropped image comprises using a saliency map to identify the updated region of interest.

5. The method of claim 1 , wherein determining a region of interest of an image based upon the eye tracking comprises identifying the region of interest using a predetermined angle associated with a field of view of the head mounted device.

6. The method of claim 1 , wherein generating a bounding box comprises using detected depths of objects in the image.

7. The method of claim 6 , wherein a center the bounding box is determined based upon a center of the region of interest and a size of a bounding box is determined based upon the detected depths of objects in the image.

8. A method of analyzing objects in images recorded by a camera of a head mounted device, the method comprising:

performing eye tracking while recording the images;

determining a region of interest of an image based upon the eye tracking;

generating a bounding box based upon a center of the region of interest and depth information associated with objects in the image, wherein a size of the bounding box is determined by the depth information;

cropping an image based upon the bounding box to generate a cropped image; and

detecting an object in the cropped image.

9. The method of claim 8 , wherein detecting an object in the cropped image comprises performing deep learning.

10. The method of claim 9 , further comprising generating an object detection confidence value associated with an object in the image, and using the object detection confidence value to determine a region of interest of a later image.

11. The method of claim 8 , further comprising performing fine cropping of the cropped image using a saliency map to identify an updated region of interest.

12. The method of claim 8 , further comprising detecting depths of objects in the image.

13. An electronic device for analyzing objects in images recorded by a camera of a head mounted device, the electronic device comprising:

a processor coupled to receive images captured by the camera, wherein the processor:

performs eye tracking while recording the images;

determines a region of interest of an image based upon the eye tracking;

generates a bounding box based upon the region of interest;

crops an image based upon the bounding box to generate a cropped image;

performs a fine cropping of the cropped image to generate an updated region of interest; and

detects an object in the cropped image after the fine cropping.

14. The electronic device of claim 13 , wherein detecting an object in the cropped image comprises performing deep learning.

15. The electronic device of claim 14 , wherein the processor further generates an object detection confidence value associated with the object in the image, and uses the object detection confidence value to determine a region of interest of a later image.

16. The electronic device of claim 13 , wherein performing a fine cropping of the cropped image comprises using a saliency map to identify the updated region of interest.

17. The electronic device of claim 13 , wherein determining a region of interest of an image based upon the eye tracking comprises identifying the region of interest using a predetermined angle associated with a field of view of the head mounted device.

18. The electronic device of claim 13 , wherein generating a bounding box comprises using detected depths of objects in the image.

19. The electronic device of claim 18 , wherein the processor further determines a center the bounding box based upon a center of the region of interest and determines a size of a bounding box based upon the detected depths of objections in the image..