US 12,272,148 B2
Scalable semantic image retrieval with deep template matching
Donna Roy, Santa Clara, CA (US); Suraj Kothawade, Dallas, TX (US); Elmar Haussmann, Stuttgart (DE); Jose Manuel Alvarez Lopez, Mountain View, CA (US); Michele Fenzi, Munich (DE); and Christoph Angerer, Munich (DE)
Assigned to Nvidia Corporation, Santa Clara, CA (US)
Filed by Nvidia Corporation, Santa Clara, CA (US)
Filed on Apr. 9, 2021, as Appl. No. 17/226,584.
Claims priority of provisional application 63/111,559, filed on Nov. 9, 2020.
Prior Publication US 2022/0147743 A1, May 12, 2022
Int. Cl. G06V 20/56 (2022.01); G06F 18/2113 (2023.01); G06F 18/214 (2023.01); G06F 18/22 (2023.01); G06N 3/08 (2023.01); G06V 30/262 (2022.01)
CPC G06V 20/56 (2022.01) [G06F 18/2113 (2023.01); G06F 18/2155 (2023.01); G06F 18/22 (2023.01); G06N 3/08 (2013.01); G06V 30/274 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
processing a first image using an object detector to identify a plurality of semantic features;
identifying a subset of the plurality of semantic features in a region of the first image corresponding to a first object;
determining similarity values between the subset of semantic features and semantic features of a plurality of unlabeled images using, at least in part, one or more similarity tensors; and
selecting, based at least in part upon the similarity values, one or more of the unlabeled images as including a representation of at least one additional object that is semantically similar to the first object.