US 11,790,568 B2
Image entity extraction and granular interactivity articulation
Tiberiu Suto, Franklin, NY (US); Shikhar Kwatra, San Jose, CA (US); Vijay Ekambaram, Chennai (IN); and Hemant Kumar Sivaswamy, Pune (IN)
Assigned to KYNDRYL, INC, New York, NY (US)
Filed by KYNDRYL, INC., New York, NY (US)
Filed on Mar. 29, 2021, as Appl. No. 17/215,432.
Prior Publication US 2022/0309715 A1, Sep. 29, 2022
Int. Cl. G06T 11/00 (2006.01); G06F 3/048 (2013.01); G06F 16/51 (2019.01); G06F 40/295 (2020.01); G06V 10/22 (2022.01); G06V 20/20 (2022.01)
CPC G06T 11/00 (2013.01) [G06F 16/51 (2019.01); G06F 40/295 (2020.01); G06V 10/22 (2022.01); G06V 20/20 (2022.01)] 19 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
identifying two or more distinct elements in an image;
generating sub-images for the two or more distinct elements;
for the distinct elements, generating adjectives descriptive of preexisting content associated with the respective distinct element by extracting the adjectives from the preexisting content;
detecting a user interaction with a particular one of the sub-images;
displaying a response list of responses including the adjectives associated with the particular element of the sub-image being interacted with in response to the user interaction with the sub-image, the list including a response having custom input field for enabling user input of a customized response;
obtaining annotation data based at least in part on user selection of one of the responses from the response list displayed for the distinct element of the sub-image being interacted with; and
assigning the annotation data to the distinct element of the sub-image being interacted with, wherein the annotation data is displayed in response to subsequent interactions with the distinct element in the image;
obtaining comment data representative of one or more comments associated with the image from an internet-based platform outputting the image, the one or more comments being posted by one or more previous users;
identifying named objects included in the one or more comments;
generating embedding representations for the named objects identified;
generating embedding representations for the two or more distinct elements identified in the image;
determining embedding similarities for the named object embedding representations and the distinct element embedding representations;
determining one or more suitable comments for a particular one of the distinct elements based, at least in part, on the embedding similarity thereof; and
providing the one or more suitable comments to be overlaid on the particular distinct element in the image in response to an interaction with the image.