System and method for automated object detection in an image
A contour/shape detection model may use relatively simple and efficient kernels to detect target edges in an object within an image or video. A co-occurrence probability may be calculated for two or more edge features in an image or video using an object definition. Edge features may be differentiated between in response to measured contextual support, and prominent edge features may be extracted based on the measured contextual support. The object may then be identified based on the extracted prominent edge features.
1. A computer-implemented method, comprising:
calculating, by a computing system, a co-occurrence probability for two or more edge features in an image or video using an object definition;
differentiating between edge features, by the computing system, in response to a measured contextual support;
extracting, by the computing system, prominent edge features based on the measured contextual support; and
identifying the object, by the computing system, based on the extracted prominent edge features.
2. The computer-implemented method of claim 1 , further comprising:
receiving, by the computing system, a request to identify an object in the image or video, wherein
the request comprises one or more of inputting, capturing, determining, calculating, and receiving input from a user.
3. The computer-implemented method of claim 2 , wherein the request comprises classes, sets, subsets, or groupings of objects.
4. The computer-implemented method of claim 1 , wherein the calculating of the co-occurrence probability for the two or more edge features further comprises one or more of computing, determining, generating, and calculating an n-wise co-occurrence probability of oriented edges conditioned on a presence or absence of globally salient contours.
5. The computer-implemented method of claim 1 , wherein the object definition comprises one or more images or representations of the object spanning a desired range of poses and points of view.
6. The computer-implemented method of claim 1 , wherein the object definition is derived from a learned set of kernels that utilizes a training set or the object definition is derived from a training corpus including both target images and distractor images.
7. The computer-implemented method of claim 6 wherein, when a training corpus is used, the training corpus is used, in part, to calculate the co-occurrence probability by determining and/or calculating a probability of two edge features being part of a single object/distractor.
8. The computer-implemented method of claim 1 , wherein the differentiation comprises one or more of separating, categorizing, distinguishing, and differentiating two or more edge features into two or more categories.
9. The computer-implemented method of claim 1 , further comprising:
constructing, by the computing system, an object-distractor difference (ODD) kernel; and
convolving the ODD kernel, by the computing system, with every edge feature, or a majority of edge features, in the image or video to obtain lateral contextual support at each location and/or orientation across the entire image or video.
10. The computer-implemented method of claim 9 , wherein the ODD kernel comprises a plurality of separate kernels having unique or quasi-unique orientations against which edge elements are measured.
11. The computer-implemented method of claim 1 , wherein the extracting of the prominent edge features comprises one or more of preserving, identifying, sequestering, promoting, and extracting edge features that have at least a predetermined level of contextual support through an object-distractor difference (ODD) kernel while one or more of suppressing, eliminating, and demoting edge features that do not have at least the predetermined level of contextual support.
12. The computer-implemented method of claim 1 , wherein n-wise, hierarchical ODD kernels are used that detect and segment one object category from everything else.
13. A computer program embodied on a non-transitory computer-readable medium, the computer program configured to cause at least one processor to:
generate a training corpus that is divided into target and distractor images or videos;
calculate estimates of n-wise co-occurrence probabilities of oriented edges conditioned on presence or absence of globally salient contours; and
generate a kernel in response to each calculated n-wise co-occurrence probability value.
14. The computer program of claim 13 , wherein the program is further configured to cause the at least one processor to:
determine differences between probability distributions determined when generating the kernels by computing a relative frequency with which a given edge feature occurs in associating with one or more surrounding edge features when the given edge feature either is, or is not, a component of a target object; and
generate object-distractor difference (ODD) kernels from the determined differences in the probability distributions.
15. The computer program of claim 14 , wherein the program is further configured to cause the at least one processor to convolve the ODD kernels with each edge feature in an image or video to obtain lateral contextual support at each location and orientation across the entire image or video.
16. The computer program of claim 15 , wherein the program is further configured to cause the at least one processor to preserve edge features with at least a predetermined level of contextual support from surrounding edges.
17. The computer program of claim 16 , wherein the program is further configured to cause the at least one processor to suppress edge features with less than the predetermined level of contextual support from the surrounding edges.
18. The computer program of claim 17 , wherein the program is further configured to cause the at least one processor to identify an object in the image or video based on the preserved edge features.
19. A system, comprising:
an image provider configured to distribute, provide, present, capture, record, process, and/or transmit one or more images and/or videos for object identification; and
an edge detection system configured to capture, identify, detect, and/or extract a portion of an image or video from a larger image or video provided by the image provider based on contour and shape, the edge detection system configured to generate object-distractor difference (ODD) kernels that utilize co-occurrence statistics of n-tuples of edge features and convolve the ODD kernels with every edge feature in the image or video to obtain lateral contextual support at each location and orientation across the image or video.
20. The system of claim 19 , further comprising:
a user device configured to receive and transmit user input relating to an object identification and/or classification, receive, from the edge detection system, a determination of the object identification in an image or video, and present the object identification to the user.