公开(公告)号：US12099913B2

公开(公告)日：2024-09-24

申请号：US16698711

申请日：2019-11-27

Applicant: ELECTRONICS AND TELECOMMUNICATIONS RESEARCH INSTITUTE

Inventor： Hye-Jin Kim ,  Sang-Yun Oh ,  Jong-Eun Lee

IPC: G06N3/04 ,  G06N3/082

CPC classification number: G06N3/04 ,  G06N3/082

Abstract: Disclosed herein are a neural-network-lightening method using a repetition-reduction block and an apparatus for the same. The neural-network-lightening method includes stacking (accumulating) an output layer (value) of either one or both of a layer constituting a neural network and a repetition-reduction block in a Condensed Decoding Connection (CDC), and lightening the neural network by reducing a feature map, generated to correspond to data stacked in the CDC, based on a preset reduction layer.

公开(公告)号：US12112468B2

公开(公告)日：2024-10-08

申请号：US17163051

申请日：2021-01-29

Applicant: Electronics and Telecommunications Research Institute

Inventor： Hye-Jin Kim ,  Suyoung Chi

IPC: G06K9/00 ,  G06N20/00 ,  G06T7/00 ,  G06T7/50

CPC classification number: G06T7/0006 ,  G06N20/00 ,  G06T7/50 ,  G06T2207/20081 ,  G06T2207/30108

Abstract: An apparatus for detecting a dimension error obtains an image of a target object, estimates dimensional data for a region of interest (ROI) for which dimensions are to be measured from the image of the target object using a learned dimensional measurement model, and determines whether there is a dimension error in the ROI from the estimated dimension data using a learned dimension error determination model.

公开(公告)号：US10043285B2

公开(公告)日：2018-08-07

申请号：US15256447

申请日：2016-09-02

Applicant: ELECTRONICS AND TELECOMMUNICATIONS RESEARCH INSTITUTE

Inventor： Hye-Jin Kim

IPC: G06K9/00 ,  G06T7/529 ,  G06K9/62

Abstract: The disclosure relates to a method and an apparatus for extracting depth information from an image. A method for extracting depth information based on machine learning according to an exemplary embodiment of the present disclosure includes generating a depth information model corresponding to at least one learning image by performing machine learning using the at least one learning image and a plurality of depth information corresponding to the at least one learning image; and extracting depth information of a target image by applying the depth information model into the target image. Embodiments of the disclosure may allow extracting precise depth information from a target image.