Evaluation of three-dimensional scenes using two-dimensional representations

• US 9,111,375 B2
• Filed: 01/07/2013
• Issued: 08/18/2015
• Est. Priority Date: 01/05/2012
• Status: Active Grant

• Current Assignee: Philip Meier
• Original Assignee: Philip Meier
• Inventors: Meier, Philip
• Primary Examiner(s): MARIAM, DANIEL G
• Application Number:
  US13/736,060
• Publication Number:
  US 20130177235A1
• Time in Patent Office: 953 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G06T 15/00 3D [Three Dimensional] image rendering
• CPC Class Codes: G06T 19/003 Navigation within 3D models or images
• CPC Class Codes: G06V 10/772 Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries

Abstract

A system adapted to implement a learning rule in a three-dimensional (3D) environment is described. The system includes: a renderer adapted to generate a two-dimensional (2D) image based at least partly on a 3D scene; a computational element adapted to generate a set of appearance features based at least partly on the 2D image; and an attribute classifier adapted to generate at least one set of learned features based at least partly on the set of appearance features and to generate a set of estimated scene features based at least partly on the set of learned features. A method labels each image from among the set of 2D images with scene information regarding the 3D scene; selects a set of learning modifiers based at least partly on the labeling of at least two images; and updates a set of weights based at least partly on the set of learning modifiers.

Claims

1. A system disposed to implement a learning rule in a three-dimensional (3D) environment, the system comprising:

a renderer disposed to generate a plurality of two-dimensional (2D) images based at least partly on a set of 3D scenes;

a computational element disposed to generate a set of appearance features based at least partly on at least one 2D image from among the plurality of 2D images; and

an attribute classifier disposed to generate at least one set of learned features based at least partly on each set of appearance features and to generate a set of estimated scene features based at least partly on each set of learned features.

2. The system of claim 1 further comprising a plurality of update elements, including a first update element disposed to evaluate the set of estimated scene features and the learning rule and update a first set of parameters used to generate a particular set of learned features from among the at least one set of learned features.

3. The system of claim 2 further including a second update element disposed to evaluate the set of estimated scene features and the learning rule and update a second set of parameters used to generate the set of estimated scene features.

4. The system of claim 3, wherein at least one 3D scene comprises a virtual environment.

5. The system of claim 3, wherein the renderer is a camera disposed to capture each 2D image.

6. The system of claim 3, wherein each 3D scene comprises a set of true labels of scene features, and each of the first and second sets of updated parameters is based at least partly on the set of true labels.

7. The system of claim 6, wherein each set of true labels comprises a spatial map of Boolean values.

8. The system of claim 3, wherein:

each learned feature in the set of learned features is calculated based at least partly on a non-linear function applied to a sum of cross products of a vector of appearance features and a 2D update matrix updated by the first update element; and

each estimated scene feature in the set of estimated scene features is calculated based at least partly on the non-linear function applied to a sum of cross products of a vector of learned features and a 2D update matrix updated by the second update element.

9. An automated method disposed to provide learning about a three-dimensional (3D) scene using a set of two-dimensional (2D) images, the method comprising:

labeling each image from among the set of 2D images with scene information regarding the 3D scene, the scene information being associated with the labeled image;

selecting a set of learning modifiers based at least partly on the labeling of at least two images; and

updating a set of weights based at least partly on the set of learning modifiers.

10. The automated method of claim 9 further comprising determining whether the selected set of modifiers is appropriate for a particular image.

11. The automated method of claim 9 further comprising determining whether to render and learn from a particular 3D scene.

12. The automated method of claim 9, wherein the 3D scene includes a set of true labels regarding object properties associated with the 3D scene.

13. The automated method of claim 9, further comprising rendering and sampling each 2D image from the set of 2D images.

14. The automated method of claim 9 further comprising evaluating at least two 2D images from the set of 2D images using a current visual algorithm.

15. The automated method of claim 9, wherein the labeling, selecting, and updating operations are performed iteratively for each image in the set of 2D images.

16. The automated method of claim 9, wherein each image in the set of 2D images is associated with the 3D scene.

17. A non-transitory computer readable medium storing an image evaluation application disposed to provide learning about a three-dimensional (3D) scene using a set of two-dimensional (2D) images, the application comprising sets of instructions for:

labeling each image from among the set of 2D images with scene information regarding the 3D scene, the scene information being associated with the labeled image;

selecting a set of learning modifiers based at least partly on the labeling of at least two images; and

updating a set of weights based at least partly on the set of learning modifiers.

18. The non-transitory computer readable medium of claim 17 further comprising sets of instructions for determining whether the selected set of modifiers is appropriate for a particular image.

19. The non-transitory computer readable medium of claim 17 further comprising sets of instructions for determining whether to render and learn from a particular 3D scene.

20. The non-transitory computer readable medium of claim 17, wherein the 3D scene includes a set of true labels regarding object properties associated with the 3D scene.