US 11,790,646 B2
Network for interacted object localization
Mert Kilickaya, Amsterdam (NL); and Arnold Wilhelmus Maria Smeulders, Amsterdam (NL)
Assigned to QUALCOMM Technologies, Inc., San Diego, CA (US)
Filed by QUALCOMM Technologies, Inc., San Diego, CA (US)
Filed on Jun. 25, 2021, as Appl. No. 17/359,379.
Prior Publication US 2022/0414371 A1, Dec. 29, 2022
Int. Cl. G06V 20/00 (2022.01); G06V 40/10 (2022.01); G06V 10/75 (2022.01); G06F 18/241 (2023.01); G06F 18/213 (2023.01)
CPC G06V 20/00 (2022.01) [G06F 18/213 (2023.01); G06F 18/241 (2023.01); G06V 10/751 (2022.01); G06V 40/10 (2022.01)] 28 Claims
OG exemplary drawing
 
1. A method, comprising:
receiving an image;
extracting a first set of human-object features from multiple positions of the image based on a learned embedding of fixed positional information;
predicting human-object pairs based on the extracted first set of human-object features; and
determining a human-object interaction based on a set of candidate interactions and the predicted human-object pairs, using a corresponding scoring matrix, the scoring matrix indicating an alignment between a candidate interaction of the set of candidate interactions and a human-object pair of the predicted human object pairs.