US 12,013,700 B2
Unsupervised contrastive learning of visual representations using negative mixing
Ioannis Kalantidis, Grenoble (FR); Diane Larlus, La Tronche (FR); Philippe Weinzaepfel, Montbonnot-Saint-Martin (FR); Mert Bulent Sariyildiz, Grenoble (FR); and Noé Pion, Plaisance (FR)
Assigned to NAVER CORPORATION, Gyeonggi-Do (KR)
Filed by NAVER CORPORATION, Gyeonggi-do (KR)
Filed on Apr. 23, 2021, as Appl. No. 17/239,501.
Claims priority of provisional application 63/086,645, filed on Oct. 2, 2020.
Prior Publication US 2022/0107645 A1, Apr. 7, 2022
Int. Cl. G05D 1/00 (2006.01); G06F 16/532 (2019.01); G06F 16/55 (2019.01); G06F 16/56 (2019.01); G06F 18/214 (2023.01); G06F 18/22 (2023.01); G06F 18/241 (2023.01); G06N 5/04 (2023.01); G06N 20/00 (2019.01); G06T 9/00 (2006.01); G06V 20/56 (2022.01)
CPC G05D 1/0212 (2013.01) [G05D 1/0231 (2013.01); G06F 16/532 (2019.01); G06F 16/55 (2019.01); G06F 16/56 (2019.01); G06N 5/04 (2013.01); G06N 20/00 (2019.01); G06T 9/002 (2013.01); G06F 18/2155 (2023.01); G06F 18/22 (2023.01); G06F 18/241 (2023.01); G06V 20/56 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A training system comprising:
an encoder module configured to receive a query image and to generate a first vector representative of one or more features in the query image using an encoder;
a mixing module configured to generate a second vector by mixing a third vector, representative of one or more features in a second image that is classified as a negative relative to the query image, with a fourth vector; and
an adjustment module configured to train the encoder by selectively adjusting one or more parameters of the encoder based on the first vector and the second vector.