Accelerating object detection

• US 9,471,828 B2
• Filed: 07/28/2014
• Issued: 10/18/2016
• 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:
  US14/444,560
• Publication Number:
  US 20160027181A1
• Time in Patent Office: 813 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
• 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:

receiving an image;

extracting coarse features from the image;

generating an object probability map based on the coarse features extracted from the image, 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 to enable detection of one or more objects in the image, wherein the dense features are features extracted using smaller sampling steps than sampling steps used to extract the coarse features.

2. The computer-implemented method of claim 1, wherein the object probability map further indicates low-probability object regions in the image that are not likely to contain an object, and wherein dense features are not extracted from the low-probability object regions of the image.

3. The computer-implemented method of claim 1, further comprising detecting one or more objects in the image based on the dense features extracted from the image.

4. The computer-implemented method of claim 1, wherein the generating the object probability map comprises:

quantizing the extracted features into visual words;

locating corresponding visual words from a feature vocabulary, the feature vocabulary associating visual words with corresponding importance weights;

assigning the corresponding importance weights of the corresponding visual word to the coarse features extracted from the image;

accumulating the importance weights of the coarse features to generate probability scores; and

assigning the probability scores to positions on the object probability map that correspond to respective locations of the coarse features in the image.

5. The computer-implemented method of claim 4, wherein the extracting dense features from the high-probability object regions of the image further comprises:

comparing the probability score of each dense feature to a predetermined threshold; and

extracting the dense feature if the probability score is greater than the threshold.

6. The computer-implemented method of claim 1, further comprising:

generating a new object probability map based on the dense features, the new object probability map indicating additional high-probability object regions in the image; and

extracting further dense features from the additional high-probability object regions of the image.

7. The computer-implemented method of claim 1, wherein the one or more objects each comprise a face.

8. 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;

receiving an image;

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;

applying a first subset of exemplars to the dense features to detect object candidate regions;

extracting additional features from the object candidate regions; and

applying one or more additional subsets of exemplars to the additional features extracted from the object candidate regions to validate one or more of the object candidate regions.

9. The system of claim 8, wherein the object probability map further indicates low-probability object regions in the image that are not likely to contain an object, and wherein dense features are not extracted from the low-probability object regions of the image.

10. The system of claim 8, wherein the generating the object probability map further comprises:

quantizing the extracted features into visual words;

locating corresponding visual words from a feature vocabulary, the feature vocabulary associating visual words with corresponding importance weights;

assigning the corresponding importance weights of the corresponding visual word to the coarse features extracted from the image;

accumulating the importance weights of the coarse features to generate probability scores; and

assigning the probability scores to positions on the object probability map that correspond to respective locations of the coarse features in the image.

11. The system of claim 10, wherein the extracting dense features from the high-probability object regions of the image further comprises:

comparing the probability score of each dense feature to a predetermined threshold; and

extracting the dense feature if the probability score is greater than the threshold.

12. The system of claim 8, wherein the number of exemplars in the first subset of exemplars is less than the number of exemplars in each of the one or more additional subsets of exemplars.

13. The system of claim 8, wherein validating the object candidate regions further comprises detecting at least one of the object candidate regions as a false positive region, and removing the false positive region from the object candidate regions.

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

an adaptive sampling module to;

extract coarse features from an image;

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

extract dense features from the high-probability object regions of the image identified in the object probability map to enable detection of one or more objects in the image, wherein the dense features are features extracted using smaller sampling steps than sampling steps used to extract the coarse features.

15. The system of claim 14, wherein the object probability map further indicates low-probability object regions in the image that are not likely to contain an object, and wherein dense features are not extracted from the low-probability object regions of the image.

16. The system of claim 14, wherein the adaptive sampling module is further configured to detect one or more objects in the image based on the dense features extracted from the image.

17. The system of claim 14, wherein the adaptive sampling module is configured to generate the object probability map by:

quantizing the extracted features into visual words;

locating corresponding visual words from a feature vocabulary, the feature vocabulary associating visual words with corresponding importance weights;

assigning the corresponding importance weights of the corresponding visual word to the coarse features extracted from the image; and

accumulating the importance weights of the coarse features to generate probability scores; and

assigning the probability scores to positions on the object probability map that correspond to respective locations of the coarse features in the image.

18. The system of claim 14, wherein the adaptive sampling module is configured to extract dense features from the high-probability object regions of the image further by:

comparing the probability score of each dense feature to a predetermined threshold; and

extracting the dense feature if the probability score is greater than the threshold.

19. The system of claim 14, wherein the adaptive sampling module is further configured to:

generate a new object probability map based on the dense features, the new object probability map indicating additional high-probability object regions in the image; and

extract further dense features from the additional high-probability object regions of the image.

20. The system of claim 14, wherein the one or more objects each comprise a face.