公开(公告)号：US11494885B2

公开(公告)日：2022-11-08

申请号：US17040308

申请日：2019-03-28

Applicant: Samsung Electronics Co., Ltd.

Inventor： Nari Im ,  Ildo Kim ,  Jaegon Kim ,  Sangjin Lee ,  Yongju Lee ,  Changgwun Lee ,  Seunghye Chyung ,  Hyeyun Jung ,  Jiyoon Park ,  Jonghoon Won

IPC: H04N5/228 ,  G06T5/50 ,  G06T7/33 ,  H04N5/265 ,  G06V10/60

Abstract: An electronic device comprises a first camera capable of photographing a designated direction, a second camera capable of photographing a direction different from the designated direction, a memory, and a processor, wherein the processor can be configured to acquire first images by using the first camera and second images by using the second camera, confirm at least one reflective object having a reflective attribute related to the light reflection in the first images, and synthesize, on the basis of the reflective attribute, at least a part of the second images with at least a part of an area corresponding to the at least one reflective object.

公开(公告)号：US20230004816A1

公开(公告)日：2023-01-05

申请号：US17716292

申请日：2022-04-08

Applicant: SAMSUNG ELECTRONICS CO., LTD.

Inventor： Changgwun Lee ,  Kyoungyoung Kim ,  Byeoungsu Kim ,  Jaegon Kim ,  Hanyoung Yim ,  Jungmin Choi ,  Sanghyuck Ha

IPC: G06N3/10 ,  G06N3/04

Abstract: In a method of optimizing a neural network model, first model information about a first neural network model is received. Device information about a first target device that is used to execute the first neural network model is received. An analysis whether the first neural network model is suitable for executing on the first target device is performed, based on the first model information, the device information, and at least one of a plurality of suitability determination algorithms. A result of the analysis is output such that the first model information and the result of the analysis are displayed on a screen.

公开(公告)号：US20220335293A1

公开(公告)日：2022-10-20

申请号：US17672204

申请日：2022-02-15

Applicant: SAMSUNG ELECTRONICS CO., LTD.

Inventor： Jungmin Choi ,  Kyoungyoung Kim ,  Byeoungsu Kim ,  Jaegon Kim ,  Changgwun Lee ,  Hanyoung Yim ,  Sanghyuck Ha

IPC: G06N3/08

Abstract: A method of optimizing a neural network model includes receiving original model information about a first neural network model that is pre-trained; generating a second neural network model and compressed model information about the second neural network model by performing a compression on the first neural network model; and outputting, on a screen, at least a part of the original model information and at least a part of the compressed model information.

公开(公告)号：US20210264260A1

公开(公告)日：2021-08-26

申请号：US17033088

申请日：2020-09-25

Applicant: SAMSUNG ELECTRONICS CO., LTD.

Inventor： BYEOUNGSU KIM ,  Kyoungyoung Kim ,  Jaegon Kim ,  Changgwun Lee ,  Sanghyuck Ha

IPC: G06N3/08 ,  G06N3/04

Abstract: The present disclosure relates to neural network training. The neural network training relates to a training method, a training device, and a system including the neural network. The neural network training includes extracting annotation data and first reliability values for first data using a neural network trained based on training data, selecting second data from among the first data based on the second data having second reliability values greater than or equal to a threshold value, expanding the training data based on the second data, and retraining the neural network based on the expanded training data

公开(公告)号：US11425300B2

公开(公告)日：2022-08-23

申请号：US16970259

申请日：2019-01-31

Applicant: SAMSUNG ELECTRONICS CO., LTD.

Inventor： Nari Im ,  Ildo Kim ,  Jaegon Kim ,  Sangjin Lee ,  Yongju Lee ,  Changgwun Lee ,  Seunghye Chyung ,  Hyeyun Jung ,  Jiyoon Park ,  Jonghun Won

IPC: H04N5/232 ,  H04N5/38

Abstract: According to various embodiments, an electronic device comprises a communication module, a memory, and a processor functionally connected with the communication module and the memory, wherein the processor may be configured to: receive a request for revision information corresponding to a first image from an external electronic device via the communication module; generate, on the basis of image recognition, first revision information by using a raw image corresponding to the first image; determine, on the basis of a comparison between the first image and the first revision information, whether a feature of the first revision information satisfies a predetermined condition or not; transmit, when the feature of the first revision information does not satisfy the predetermined condition, the first revision information to the external device to enable the external electronic device to generate a second image, that is, the revised first image by using the first revision information; generate, when the feature of the first revision information satisfies the predetermined condition, second revision information having a feature which is at least partially different from the first revision information, by using the first image and the first revision information; and transmit the second revision information to the external device to allow the external electronic device to generate a third image, that is, the revised first image by using the second revision information. Other embodiments are possible.

公开(公告)号：US11270420B2

公开(公告)日：2022-03-08

申请号：US16651997

申请日：2018-09-20

Applicant: Samsung Electronics Co., Ltd.

Inventor： Changgwun Lee ,  Ildo Kim ,  Jaeho Lee ,  Hyeyun Jung ,  Kihuk Lee

IPC: G06T5/50 ,  G06F3/0482 ,  G06K9/62

Abstract: Various embodiments provide an electronic device and a method, the electronic device comprising a communication module, a memory, and a processor, wherein the processor is configured to: recognize at least one object from among one or more objects by using an image containing the one or more objects; identify a recognition rate and a category corresponding to the at least one object at least on the basis of the recognition; obtain at least one reference image corresponding to the object at least on the basis of the category; when the recognition rate satisfies a first specified condition associated with the recognition rate corresponding to the category, correct the at least one object or an area corresponding to the at least one object by using a reference image, which satisfies the first specified condition, from among the at least one reference image; and when the recognition rate satisfies a second specified condition associated with the recognition rate corresponding to the category, correct the at least one object or the area corresponding to the at least one object by using a reference image selected in accordance with an input from among the at least one reference image. In addition, other embodiments are also possible.