| CPC G06T 7/74 (2017.01) [G06T 2207/30196 (2013.01)] | 16 Claims |
|
1. An apparatus, comprising:
one or more processors configured to:
obtain a first set of images of a person, wherein each of the first set of images is captured by a respective image sensor;
determine respective two-dimensional (2D) representations of a keypoint of the person based on the first set of images, wherein each 2D representation is associated with a confidence score;
determine a first three-dimensional (3D) representation of the keypoint based on the 2D representations of the keypoint and the confidence score associated with each 2D representation, wherein, during the determination of the first 3D representation of the keypoint, the one or more processors are configured at least to:
arrange the 2D representations of the keypoint into multiple groups, wherein each group includes two or more of the 2D representations;
predict, for each of the multiple groups, a respective candidate 3D representation of the keypoint by triangulating the two or more 2D representations of the keypoint included in the group;
calculate a respective prediction error associated with the prediction of the candidate 3D representation for each of the multiple groups, wherein the calculation is made at least by obtaining 2D re-projections of the candidate 3D representation in respective coordinate systems associated with the two or more 2D representations of the group and calculating a sum of differences between the 2D re-projections and the two or more 2D representations; and
select the candidate 3D representation with the smallest calculated prediction error as the first 3D representation of the keypoint; and
estimate a 3D pose of the person based on at least the first 3D representation of the keypoint.
|