US 12,367,233 B2
Determining 3D models corresponding to an image
Robert Banfield, Riverview, FL (US); Aristodimos Komninos, Athens (GR); Jacques Harvent, Le Pereux-sur-Marne (FR); Michael Tadros, Boulder, CO (US); and Karolina Torttila, Helsinki (FI)
Assigned to Trimble Inc., Westminster, CO (US)
Filed by Trimble Inc., Westminster, CO (US)
Filed on Aug. 11, 2023, as Appl. No. 18/233,181.
Prior Publication US 2024/0104132 A1, Mar. 28, 2024
Int. Cl. G06F 16/00 (2019.01); G06F 16/532 (2019.01); G06F 16/538 (2019.01); G06F 16/56 (2019.01); G06T 15/10 (2011.01); G06T 17/00 (2006.01)
CPC G06F 16/532 (2019.01) [G06F 16/538 (2019.01); G06F 16/56 (2019.01); G06T 15/10 (2013.01); G06T 17/00 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
receiving, by a computing device, image data that corresponds to at least one two-dimensional image;
generating, by the computing device and using supervised layers of a hybrid machine-learning model, features based on the image data corresponding to the at least one two-dimensional image;
generating, by the computing device and by using an unsupervised layer of the hybrid machine-learning model, a representation vector for the at least one two-dimensional image by transforming the features into a predetermined amount of numerical representations corresponding to the features; and
outputting, by the computing device, the representation vector for the at least one two-dimensional image to facilitate a search query for one or more three-dimensional models associated with the at least one two-dimensional image;
wherein the hybrid machine-learning model has an architecture such that each of the supervised layers precedes the unsupervised layer and the supervised layers generate an input for the unsupervised layer, each of the supervised layers having been trained via a supervised training technique and the unsupervised layer having been trained via an unsupervised training technique.