METHOD AND SYSTEM FOR MESHING HUMAN AND COMPUTER COMPETENCIES FOR OBJECT CATEGORIZATION

• US 20090137924A1
• Filed: 01/29/2009
• Published: 05/28/2009
• Est. Priority Date: 08/27/2007
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Shenoy, Pradeep U., Horvitz, Eric J., Tan, Desney S., Kapoor, Ashish
• Granted Patent: US 8,688,208 B2
• Time in Patent Office: Days
• US Class Current: 600/545
• CPC Class Codes: A61B 5/316 Modalities, i.e. specific diagnostic methods
• CPC Class Codes: A61B 5/378 Visual stimuli
• CPC Class Codes: A61B 5/7267 involving training the classification device
• CPC Class Codes: G06F 18/23 Clustering techniques
• CPC Class Codes: G06F 18/29 Graphical models, e.g. Bayesian networks
• CPC Class Codes: G06V 10/762 using clustering, e.g. of similar faces in social networks
• CPC Class Codes: G06V 10/84 using probabilistic graphical models from image or video features, e.g. Markov models or Bayesian networks
• CPC Class Codes: G06V 30/414 Extracting the geometrical structure, e.g. layout tree; Block segmentation, e.g. bounding boxes for graphics or text
• CPC Class Codes: G16H 50/50 for simulation or modelling of medical disorders
• CPC Class Codes: G16Z 99/00 Subject matter not provided for in other main groups of this subclass

Abstract

The subject disclosure relates to a method and system for visual object categorization. The method and system include receiving human inputs including data corresponding to passive human-brain responses to visualization of images. Computer inputs are also received which include data corresponding to outputs from a computerized vision-based processing of the images. The human and computer inputs are processing so as to yield a categorization for the images as a function of the human and computer inputs.

Claims

1. A method for visual object categorization, including:

receiving at least one human input via a first receiving component, the at least one human input including data corresponding to a plurality of passive human-brain responses to visualization of a plurality of images;

receiving at least one computer input via a second receiving component, the at least one computer input including data corresponding to a plurality of outputs from a computerized vision-based processing of the plurality of images; and

processing the at least one human input and the at least one computer input with a processor, the processing step yielding a categorization for at least one of the plurality of images as a function of the at least one human input and the at least one computer input.

2. The method of claim 1, at least one of the plurality of passive human-brain responses being an electroencephalograph (EEG).

3. The method of claim 1, the at least one human input providing similarities between the passive human-brain responses for each of the plurality of images, the at least one computer input providing similarities between the outputs for each of the plurality of images.

4. The method of claim 1, the at least one computer input corresponding to outputs from a pyramid match kernel algorithm.

5. The method of claim 1, the processing step including processing the at least one human input and the at least one computer input according to a convex kernel algorithm.

6. The method of claim 1, the processing step further including voting on the categorization for the at least one of the plurality of images.

7. The method of claim 1 further comprising generating a predictive model for categorizing a subsequent set of images.

8. The method of claim 7, the generating step including generating a distributed computing algorithm for categorizing the subsequent set of images where at least one of the at least one human input or the at least one computer input includes missing data.

9. The method of claim 7 further comprising selectively sampling the subsequent set of images so as to improve the accuracy of the predictive model.

10. The method of claim 1 further comprising training a pure computer vision-based system with the at least one human input.

11. A computer readable medium comprising computer executable instructions for carrying out the method of claim 1.

12. A system for visual object categorization, including:

a first receiving component, the first receiving component configured to receive at least one human input including data corresponding to a plurality of passive human-brain responses to visualization of a plurality of images;

a second receiving component, the second receiving component configured to receive at least one computer input including data corresponding to a plurality of outputs from a computerized vision-based processing of the plurality of images;

a processor coupled to the first and second receiving components, the processor configured to process the at least one human input and the at least one computer input so as to yield a categorization for at least one of the plurality of images as a function of the at least one human input and the at least one computer input.

13. The system of claim 12, the processor further configured to generate a predictive model for categorizing a subsequent set of images.

14. The system of claim 13, the processor further configured to implement a Bayesian co-training model.

15. The system of claim 14, the processor further configured to assign proxy values to missing views.

16. The system of claim 13, the processor further configured to selectively sample the subsequent images.

17. The system of claim 12, the processor further configured to process at least one passive human-brain response received as an electroencephalograph (EEG).

18. The system of claim 12, the processor further configured to process at least one computer input corresponding to outputs from a pyramid match kernel algorithm.

19. The system of claim 12, the processor further configured to process the at least one human input and the at least one computer input according to a convex kernel algorithm.

20. A method for visual object categorization, including:

receiving at least one electroencephalograph (EEG) via a first receiving component, the at least one EEG including data corresponding to a plurality of passive human-brain responses to visualization of a plurality of images;

receiving at least one computer input via a second receiving component, the at least one computer input corresponding to outputs from a pyramid match kernel algorithm and including data corresponding to a plurality of outputs from a computerized vision-based processing of the plurality of images; and

processing the at least one EEG and the at least one computer input with a processor, the processing step yielding a categorization for at least one of the plurality of images according to a convex kernel algorithm as a function of the at least EEG and the at least one computer input.