公开(公告)号：US10152829B2

公开(公告)日：2018-12-11

申请号：US15521035

申请日：2015-10-12

Applicant: SAMSUNG ELECTRONICS CO., LTD.

Inventor： Ji-won Jeong ,  Do-kyoon Kim ,  Eun-jung Ju ,  Young-min Kwak ,  Seung-ho Shin ,  Sun-young Han

IPC: G06T15/00 ,  G06T19/20 ,  G06T7/73 ,  G06T13/40 ,  G06Q30/06

Abstract: Provided are a virtual fitting device and a virtual fitting method thereof. The virtual fitting method includes acquiring a live video by photographing a user, acquiring a virtual video which includes an avatar corresponding to the user based on a body shape and a movement of the user, determining each composition area of the live video and the virtual video by analyzing the live video and the virtual video, and outputting a virtual fitting video by composing the composition area of the live video and the composition area of the virtual video.

公开(公告)号：US11176402B2

公开(公告)日：2021-11-16

申请号：US16613627

申请日：2018-05-08

Applicant: Samsung Electronics Co., Ltd.

Inventor： Deok-ho Kim ,  Won-woo Lee ,  Jae-woong Lee ,  Bong-hoon Park ,  Gun-Ill Lee ,  Ji-won Jeong ,  Teresa Ka Ki Ng

IPC: G06K9/32 ,  G06K9/62 ,  G06K9/20

Abstract: Provided are an artificial intelligence (AI) system for simulating functions such as recognition, determination, and so forth of the human brain by using a mechanical learning algorithm like deep learning, or the like, and an application thereof. The disclosure relates to a method, performed by a device, of identifying an object, and the method includes obtaining an image including an object, extracting attribute information of the object from the image by using a plurality of layers included in a network for determining a category of the object, obtaining feature information representing the object by combining attribute information extracted from at least some layers among the plurality of layers by using at least one feature extraction layer, and identifying the object based on a result of comparing the obtained feature information with feature information of each of a plurality of previously stored object images, wherein a parameter of each of the at least one feature extraction layer is configured according to a training result based on a database including a plurality of images.