This AI paper proposes “SuperGlue”, a graphical neural network that simultaneously performs context aggregation, local feature matching and filtering for broad-based pose estimation.
Imagine you have two photos of the same scene taken from different angles. Most of the objects in both images are the same, you’re just looking at them from different angles. In computer vision, objects are assumed to have certain characteristics like edges, corners, etc. Matching these characteristics is critical for some applications. But what …
