US 11,915,520 B2
Finding similar persons in images
Saeid Motiian, San Francisco, CA (US); Zhe Lin, Fremont, CA (US); Shabnam Ghadar, Menlo Park, CA (US); and Baldo Faieta, San Francisco, CA (US)
Assigned to Adobe Inc., San Jose, CA (US)
Filed by Adobe Inc., San Jose, CA (US)
Filed on Sep. 2, 2022, as Appl. No. 17/902,349.
Application 17/902,349 is a continuation of application No. 17/207,178, filed on Mar. 19, 2021, granted, now 11,436,865.
Prior Publication US 2022/0415084 A1, Dec. 29, 2022
This patent is subject to a terminal disclaimer.
Int. Cl. G06V 40/16 (2022.01); G06V 30/194 (2022.01); G06V 40/10 (2022.01); G06F 18/00 (2023.01); G06F 18/20 (2023.01)
CPC G06V 40/172 (2022.01) [G06F 18/00 (2023.01); G06F 18/29 (2023.01); G06V 30/194 (2022.01); G06V 40/10 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
generating an image embedding for an input image by combining a first embedding corresponding to a first cropped image of the input image and a second embedding corresponding to a second cropped image of the input image, wherein a first machine learning model trained on a head image training dataset using contrastive learning generates the first embedding and a second machine learning model trained on a body image training dataset using contrastive learning generates the second embedding, wherein the head image training dataset originates from a first source and the body image training dataset originates from a second source, and wherein the first source is different from the second source; and
querying an image repository in embedding space by comparing the image embedding to a plurality of image embeddings associated with a plurality of images in the image repository to obtain one or more images based on similarity to the input image in the embedding space.