Back to results
US 2024/0311637 A1Published application
Systems and methods for virtual and augmented reality
AssigneeMagic Leap, Inc.
InventorsPaul-Edouard SARLIN, Daniel DeTONE, Tomasz Jan MALISIEWICZ, Andrew RABINOVICH
Life of this patent
Estimating the term…
Abstract
The description relates the feature matching. Our approach establishes pointwise correspondences between challenging image pairs. It takes off-the-shelf local features as input and uses an attentional graph neural network to solve an assignment optimization problem. The deep middle-end matcher acts as a middle-end and handles partial point visibility and occlusion elegantly, producing a partial assignment matrix.
Claims18
A computer system comprising:
- a computer-readable medium;
- a processor connected to the computer-readable medium;
- and
- a set of instructions on the computer-readable medium, including:
- a deep middle-end matcher architecture that is executable by the processor and includes:
- an attentional graph neural network that creates a representation;
- and
- an optimal matching layer that creates a score from the representation and finds an optimal assignment based on score.
5 more elements
A computer-implemented method that includes:
- creating, with an attentional graph neural network, a representation;
- creating, with an optimal matching layer, a score from the representation;
- and
- finding, with the optimal matching layer, an optimal assignment based on score.
1 more element
Dependencies are inferred from the claim text.