US 12,169,859 B1
Real-time interactive outfit recommendation
Alexander Lorbert, Givat Shmuel (IL); David Neiman, Tel-Aviv (IL); Arik Poznanski, Herzliya (IL); Eduard Oks, Redmond, WA (US); Megan E Anderson, Seattle, WA (US); and Layne Skullerud, La Mesa, CA (US)
Assigned to AMAZON TECHNOLOGIES, INC., Seattle, WA (US)
Filed by Amazon Technologies, Inc., Seattle, WA (US)
Filed on Nov. 23, 2021, as Appl. No. 17/533,903.
Int. Cl. G06Q 30/00 (2023.01); G06F 16/9535 (2019.01); G06Q 30/0601 (2023.01)
CPC G06Q 30/0631 (2013.01) [G06F 16/9535 (2019.01); G06Q 30/0627 (2013.01); G06Q 30/0643 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
receiving, by a machine learning model, first training data comprising ground truth outfit data, the ground truth outfit data comprising first fashion item data and second fashion item data;
inputting a first visual feature vector representing the first fashion item data into the machine learning model;
generating, by the machine learning model using the first visual feature vector, a first predicted visual feature vector for the second fashion item data;
generating an updated machine learning model by decreasing a difference between the first predicted visual feature vector and a ground truth visual feature vector associated with the second fashion item data;
causing a first graphical user interface (GUI) to be displayed, the first GUI depicting at least a first fashion item image associated with a first category;
determining a first user profile;
determining first filter settings of the first user profile;
receiving a first selection of a first graphical control on the first GUI, the first selection corresponding to a request for a recommendation from a second category;
determining a second fashion item image from among a plurality of fashion item images of the second category, wherein the second fashion item image has been determined by the updated machine learning model based at least in part on a second predicted visual feature vector generated by the updated machine learning model using a visual feature vector for the first fashion item image, and wherein the second fashion item image is selected in accordance with the first filter settings of the first user profile; and
causing the second fashion item image to be displayed on the first GUI.