Machine learning based model localization system
First Claim
1. A computer-operated method of determining an angular field of view and estimating a pose of a source imaging sensor based on at least one two-dimensional (2D) image input comprising a plurality of image pixels of an observed scene, the method comprising:
- (a) accessing an input data set comprising a 2D image data set to be analyzed and source imaging sensor parameter information, (b) executing a Machine Learning algorithm that uses said 2D image data set of the input data set to generate estimated depth values for at least a portion of image pixels output by the source imaging sensor to provide real three-dimensional (3D) image points having associated depth values,(c) in parallel with executing the Machine Learning algorithm, determining an angular field of view of the source imaging sensor based on the input data set and generating a the source imaging sensor angular field of view as output,(d) in response to the real 3D image points including the generated estimated depth values and the generated angular field of view, generating a source imaging sensor (3D) pose estimate relative to the real 3D image points in the observed scene, and(e) outputting the generated 3D pose estimate in conjunction with the estimated depth values.
2 Assignments
0 Petitions
Accused Products
Abstract
A method for deriving an image sensor'"'"'s 3D pose estimate from a 2D scene image input includes at least one Machine Learning algorithm trained a priori to generate a 3D depth map estimate from the 2D image input, which is used in conjunction with physical attributes of the source imaging device to make an accurate estimate of the imaging device 3D location and orientation relative to the 3D content of the imaged scene. The system may optionally employ additional Machine Learning algorithms to recognize objects within the scene to further infer contextual information about the scene, such as the image sensor pose estimate relative to the floor plane or the gravity vector. The resultant refined imaging device localization data can be applied to static (picture) or dynamic (video), 2D or 3D images, and is useful in many applications, most specifically for the purposes of improving the realism and accuracy of primarily static, but also dynamic Augmented Reality (AR) applications.
17 Citations
20 Claims
-
1. A computer-operated method of determining an angular field of view and estimating a pose of a source imaging sensor based on at least one two-dimensional (2D) image input comprising a plurality of image pixels of an observed scene, the method comprising:
-
(a) accessing an input data set comprising a 2D image data set to be analyzed and source imaging sensor parameter information, (b) executing a Machine Learning algorithm that uses said 2D image data set of the input data set to generate estimated depth values for at least a portion of image pixels output by the source imaging sensor to provide real three-dimensional (3D) image points having associated depth values, (c) in parallel with executing the Machine Learning algorithm, determining an angular field of view of the source imaging sensor based on the input data set and generating a the source imaging sensor angular field of view as output, (d) in response to the real 3D image points including the generated estimated depth values and the generated angular field of view, generating a source imaging sensor (3D) pose estimate relative to the real 3D image points in the observed scene, and (e) outputting the generated 3D pose estimate in conjunction with the estimated depth values. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)
-
-
16. A computer-operated system for determining an angular field of view and estimating a three-dimensional (3D) pose of a source imaging sensor, comprising at least one processor configured to perform operations comprising:
-
(a) receiving, based on image capture by a source imaging sensor, a two-dimensional (2D) input image data set comprising image pixels, (b) determining an angular field of view of the source imaging sensor based on the 2D input image data set, (c) using a Machine Learning algorithm that analyzes the 2D input image data set, estimating depth values for at least some of the image pixels to provide real 3D image points, and (d) in response to the determined angular field of view and the real 3D image points including the estimated depth values, estimating a 3D pose of the source imaging sensor relative to the real 3D image points. - View Dependent Claims (17, 18, 19, 20)
-
Specification