Snow classifier context window reduction using class t-scores and mean differences

• US 9,195,908 B2
• Filed: 05/22/2013
• Issued: 11/24/2015
• Est. Priority Date: 05/22/2013
• Status: Active Grant

• Current Assignee: Conduent Business Services, LLC (Conduent, Inc.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Schweid, Stuart, Burry, Aaron Michael, Kozitsky, Vladimir
• Primary Examiner(s): CHEN, ALAN S
• Application Number:
  US13/899,705
• Publication Number:
  US 20140348391A1
• Time in Patent Office: 916 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 18/2113 by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation

Abstract

Methods, systems and processor-readable media for determining, post training, which locations of a classifier window are most significant in discriminating between class and non-class objects. The important locations can be determined by calculating the mean and standard deviation of every pixel location in the classifier context for both the positive and negative samples of the classifier. Using a combination of t-scores and mean differences, the importance of all pixel locations in the classifier score can be rank ordered. A sufficient number of pixel locations can then be selected to achieve a detection rate close enough to the full classifier for a particular application.

Claims

1. A method for determining important locations of a classifier, said method comprising:

calculating a mean deviation and a standard deviation of every pixel location in a classifier with respect to positive and negative samples associated with said classifier;

ranking an importance of all pixel locations calculated with respect to said classifier based on a combination of t-scores and mean differences; and

selecting a number of said locations to achieve a detection rate for said classifier for a particular application in order to determine which locations of said classifier are involved in discrimination between class objects and non-class objects.

2. The method of claim 1 further comprising determining at least one t-score among said t-scores according to an equation:

t_loc=|m_pos−

m_neg|/(s²_pos+s²_neg)^0.5 wherein m is a mean of weights of an example set at a particular location and s² is a variance of weights of said example set.

3. The method of claim 2 wherein said example set comprises positive values or negative values.

4. The method of claim 1 wherein at least one mean difference among said mean differences comprises a mean difference between said class objects and said non-class objects.

5. The method of claim 4 further comprising calculating said at least one mean difference for a particular location according to an equation:

meandiff_loc=|m_pos−

m_neg|.

6. The method of claim 1 wherein said classifier comprises a SNoW classifier.

7. The method of claim 6 wherein said SNoW classifier is utilized in a high-speed/high-bandwidth application.

8. A system for determining important locations of a classifier, said system comprising:

a processor;

a data bus coupled to said processor; and

a computer-usable medium embodying computer program code, said computer-usable medium being coupled to said data bus, said computer program code comprising instructions executable by said processor and configured for;

calculating a mean deviation and a standard deviation of every pixel location in a classifier with respect to positive and negative samples associated with said classifier;

ranking an importance of all pixel locations calculated with respect to said classifier based on a combination of t-scores and mean differences; and

selecting a number of said locations to achieve a detection rate for said classifier for a particular application in order to determine which locations of said classifier are involved in discrimination between class objects and non-class objects.

9. The system of claim 8 wherein said instructions are further configured for determining at least one t-score among said t-scores according to an equation:

t_loc=|m_pos−

m_neg|/(s²_pos+s²_neg)^0.5 wherein m is a mean of weights of an example set at a particular location and s² is a variance of weights of said example set.

10. The system of claim 9 wherein said example set comprises positive values or negative values.

11. The system of claim 8 wherein at least one mean difference among said mean differences comprises a mean difference between said class objects and said non-class objects.

12. The system of claim 11 wherein said instructions are further configured for calculating said at least one mean difference for a particular location according to an equation:

meandiff_loc=|m_pos−

m_neg|.

13. The system of claim 8 wherein said classifier comprises a SNOW classifier.

14. The system of claim 13 wherein said SNoW classifier is utilized in a high-speed/high-bandwidth application.

15. A non-transitory processor-readable medium storing code representing instructions to cause a process to determine important locations of a classifier, said code comprising code to:

calculate a mean deviation and a standard deviation of every pixel location in a classifier with respect to positive and negative samples associated with said classifier;

rank an importance of all pixel locations calculated with respect to said classifier based on a combination of t-scores and mean differences; and

select a number of said locations to achieve a detection rate for said classifier for a particular application in order to determine which locations of said classifier are involved in discrimination between class objects and non-class objects.

16. The processor-readable medium of claim 15 wherein said code further comprises code to determine at least one t-score among said t-scores according to an equation:

t_loc=|m_pos−

m_neg|/(s²_pos+s²_neg)^0.5 where m is a mean of weights of an example set at a particular location and s² is a variance of weights of said example set.

17. The processor-readable medium of claim 16 wherein said example set comprises positive values or negative values.

18. The processor-readable medium of claim 15 wherein at least one mean difference among said mean differences comprises a mean difference between said class objects and said non-class objects.

19. The processor-readable medium of claim 18 wherein said code further comprises code to calculate said at least one mean difference for a particular location according to an equation:

meandiff_loc=|m_pos−

m_neg|.

20. The processor-readable medium of claim 15 wherein said classifier comprises a SNoW classifier and wherein said SNoW classifier is adapted for use in a high-speed/high-bandwidth application.