Accelerating object detection

• US 10,043,057 B2
• Filed: 09/01/2016
• Issued: 08/07/2018
• Est. Priority Date: 07/28/2014
• Status: Active Grant

• Current Assignee: Adobe Inc.
• Original Assignee: Adobe Systems Incorporated (Adobe Inc.)
• Inventors: Shen, Xiaohui, Lin, Zhe, Brandt, Jonathan W.
• Primary Examiner(s): Liew, Alex
• Application Number:
  US15/254,587
• Publication Number:
  US 20160371538A1
• Time in Patent Office: 705 Days
• 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
• CPC Class Codes: G06F 18/2148 characterised by the process organisation or structure, e.g. boosting cascade
• CPC Class Codes: G06V 10/464 using a plurality of salient features, e.g. bag-of-words [BoW] representations
• CPC Class Codes: G06V 10/7747 Organisation of the process, e.g. bagging or boosting
• CPC Class Codes: G06V 40/161 Detection; Localisation; Normalisation
• CPC Class Codes: G06V 40/168 Feature extraction; Face representation
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships

Abstract

Accelerating object detection techniques are described. In one or more implementations, adaptive sampling techniques are used to extract features from an image. Coarse features are extracted from the image and used to generate an object probability map. Then, dense features are extracted from high-probability object regions of the image identified in the object probability map to enable detection of an object in the image. In one or more implementations, cascade object detection techniques are used to detect an object in an image. In a first stage, exemplars in a first subset of exemplars are applied to features extracted from the multiple regions of the image to detect object candidate regions. Then, in one or more validation stages, the object candidate regions are validated by applying exemplars from the first subset of exemplars and one or more additional subsets of exemplars.

Claims

1. A computer-implemented method comprising:

selecting exemplar classifiers based on an image, the exemplar classifiers taken from example images as examples of an object that is detectable in each example image;

grouping the selected exemplar classifiers into at least a first subset of exemplar classifiers and a second subset of exemplar classifiers;

extracting features from one or more regions of the image;

applying each exemplar classifier in the first subset of exemplar classifiers to the features extracted from the one or more regions of the image to detect one or more object candidate regions;

extracting additional features from the one or more object candidate regions; and

applying each exemplar classifier in the first subset of exemplar classifiers and the second subset of exemplar classifiers to the features and the additional features extracted from the one or more object candidate regions to validate at least one of the one or more of the object candidate regions.

2. The computer-implemented method of claim 1, wherein the number of exemplar classifiers in the first subset of exemplar classifiers is less than the number of exemplar classifiers in the second subset of exemplar classifiers.

3. The computer-implemented method of claim 1, wherein the applying each exemplar classifier in the first subset of exemplar classifiers to the features extracted from the one or more regions of the image to detect the one or more object candidate regions, further comprises determining a confidence score for each exemplar classifier as applied to each of the one or more regions, the confidence score indicating whether an object is detected in the region based on the exemplar classifier.

4. The computer-implemented method of claim 3, wherein the applying each exemplar classifier in the first subset of exemplar classifiers to the features extracted from the one or more regions of the image to detect the one or more object candidate regions, further comprises:

for each of the one or more regions, summing the confidence scores of each exemplar classifier to generate an object detection score;

comparing the object detection score for each region to a predetermined threshold; and

detecting the one or more object candidate regions as regions with object detection scores above the predetermined threshold.

5. The computer-implemented method of claim 1, wherein validating the at least one of the one or more object candidate regions further comprises detecting at least one of the object candidate regions as a false positive region.

6. The computer-implemented method of claim 5, wherein validating the at least one of the one or more object candidate regions further comprises removing the false positive region from the one or more object candidate regions.

7. The computer-implemented method of claim 1, wherein applying each exemplar classifier in the first subset of exemplar classifiers and the second subset of exemplar classifiers to the features and the additional features extracted from the one or more object candidate regions to validate at least one of the one or more of the object candidate regions further comprises:

determining a confidence score for each exemplar classifier in the first group of exemplar classifiers and the second group of exemplar classifiers as applied to each of the one or more object candidate regions, the confidence score indicating whether an object is detected in the one or more object candidate region based on the exemplar classifier;

for each of the one or more object candidate regions, summing the confidence scores of each exemplar classifier to generate an object detection score;

comparing the object detection score for each of the one or more object candidate region to a predetermined threshold;

validating at least one of the one or more object candidate regions with object detection scores above the predetermined threshold; and

removing at least one of the one or more object candidate regions with object detection scores below the predetermined threshold.

8. The computer-implemented method of claim 1, wherein the additional features extracted from the one or more object candidate regions are denser than the features extracted from the one or more regions of the image.

9. The computer-implemented method of claim 1, wherein the additional features are extracted using smaller sampling steps than sampling steps used to extract the features.

10. The computer-implemented method of claim 1, wherein the extracting the features from the image further comprises:

extracting coarse features from the image;

generating an object probability map based on the extracted coarse features, the object probability map indicating high-probability object regions in the image that are likely to contain an object; and

extracting dense features from the high-probability object regions of the image identified in the object probability map.

11. The computer-implemented method of claim 1, wherein the object comprises a face.

12. A system comprising:

one or more processors;

one or more memories having instructions stored thereon that, responsive to execution by the one or more processors, perform operations comprising;

selecting exemplar classifiers based on an image, the exemplar classifiers taken from example images as examples of an object that is detectable in each example image;

grouping the selected exemplar classifiers into at least a first subset of exemplar classifiers and a second subset of exemplar classifiers;

extracting features from one or more regions of the image;

applying each exemplar classifier in the first subset of exemplar classifiers to the features extracted from the one or more regions of the image to detect one or more object candidate regions;

extracting additional features from the one or more object candidate regions; and

applying each exemplar classifier in the first subset of exemplar classifiers and the second subset of exemplar classifiers to the features and the additional features extracted from the one or more object candidate regions to validate at least one of the one or more of the object candidate regions.

13. The system of claim 12, wherein the number of exemplar classifiers in the first subset of exemplar classifiers is less than the number of exemplar classifiers in the second subset of exemplar classifiers.

14. The system of claim 12, wherein the applying each exemplar classifier in the first subset of exemplar classifiers to the features extracted from the one or more regions of the image to detect the one or more object candidate regions, further comprises:

determining a confidence score for each exemplar classifier as applied to each of the one or more regions, the confidence score indicating whether an object is detected in the region based on the exemplar classifier;

for each of the one or more regions, summing the confidence scores of each exemplar classifier to generate an object detection score;

comparing the object detection score for each region to a predetermined threshold; and

detecting the one or more object candidate regions as regions with object detection scores above the predetermined threshold.

15. The system of claim 12, wherein validating the at least one of the one or more object candidate regions further comprises detecting at least one of the one or more object candidate regions as a false positive region, and removing the false positive region from the one or more object candidate regions.

16. The system of claim 12, wherein applying each exemplar classifier in the first subset of exemplar classifiers and the second subset of exemplar classifiers to the features and the additional features extracted from the one or more object candidate regions to validate at least one of the one or more of the object candidate regions further comprises:

determining a confidence score for each exemplar classifier in the first group of exemplar classifiers and the second group of exemplar classifiers as applied to each of the one or more object candidate regions, the confidence score indicating whether an object is detected in the one or more object candidate region based on the exemplar classifier;

for each of the one or more object candidate regions, summing the confidence scores of each exemplar classifier to generate an object detection score;

comparing the object detection score for each of the one or more object candidate regions to a predetermined threshold;

validating at least one of the one or more object candidate regions with object detection scores above the predetermined threshold; and

removing at least one of the one or more object candidate regions with object detection scores below the predetermined threshold.

17. The system of claim 12, wherein the additional features extracted from the one or more object candidate regions are denser than the features extracted from the one or more regions of the image.

18. The system of claim 12, wherein the additional features are extracted using smaller sampling steps than sampling steps used to extract the features.

19. The system of claim 12, wherein the object comprises a face.

20. A system implemented at least partially in hardware, the system comprising:

a cascade detection module to;

select exemplar classifiers based on an image, the exemplar classifiers taken from example images as examples of an object that is detectable in each example image;

group the selected exemplar classifiers into at least a first subset of exemplar classifiers and a second subset of exemplar classifiers;

extract features from one or more regions of the image;

apply each exemplar classifier in the first subset of exemplar classifiers to the features extracted from the one or more regions of the image to detect one or more object candidate regions;

extract additional features from the one or more object candidate regions; and

apply each exemplar classifier in the first subset of exemplar classifiers and the second subset of exemplar classifiers to the features and the additional features extracted from the one or more object candidate regions to validate at least one of the one or more of the object candidate regions.