US 11,790,045 B2
Auto-tags with object detection and crops
Shipali Shetty, San Jose, CA (US); Zhe Lin, Bellevue, WA (US); and Alexander Smith, Orange, NH (US)
Assigned to ADOBE, INC., San Jose, CA (US)
Filed by ADOBE INC., San Jose, CA (US)
Filed on Apr. 26, 2021, as Appl. No. 17/240,246.
Prior Publication US 2022/0343108 A1, Oct. 27, 2022
Int. Cl. G06F 18/22 (2023.01); G06T 7/00 (2017.01); G06V 10/75 (2022.01)
CPC G06F 18/22 (2023.01) [G06T 7/0002 (2013.01); G06V 10/75 (2022.01); G06T 2207/20132 (2013.01); G06T 2210/12 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A method comprising:
receiving an image;
identifying an attribute tag for the image and an attribute tag confidence score for the attribute tag using an auto-tagger;
selecting the image for additional processing based on the attribute tag being among a set of problematic tags;
identifying an object in the image and an object confidence score for the object using an object detection network based on the selection, wherein the object detection network comprises an artificial neural network trained on a training set selected to include images that include objects corresponding to the set of problematic tags;
computing an updated attribute tag confidence score based on the attribute tag confidence score and the object confidence score; and
transmitting the updated attribute tag confidence score in response to receiving the image.