APPARATUS AND METHOD FOR EXTRACTING OBJECT INFORMATION

Disclosed are an object information extraction device and a method thereof. According to one embodiment of the present invention, the object information extraction device comprises: a padding image generation part for generating a padding image including an original image; a partial image acquisitio...

Full description

Saved in:
Bibliographic Details
Main Authors PARK DO YOUNG, JO SEONG HO, YANG HEE SUNG, JEON JOONG BAE
Format Patent
LanguageEnglish
Korean
Published 08.05.2020
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:Disclosed are an object information extraction device and a method thereof. According to one embodiment of the present invention, the object information extraction device comprises: a padding image generation part for generating a padding image including an original image; a partial image acquisition part for acquiring a plurality of partial images for the padding image; an object classification result acquisition part for acquiring an object classification result for each of the plurality of partial images by using an object classification model; a confidence map generation part for generating a confidence map having the same size as the padding image and including a confidence value based on the object classification result; and an object information acquisition part for acquiring object information in the padding image based on the confidence map. According to the present invention, the accuracy of object classification and object position extraction can be improved, and time and cost required for collecting learning data including position information can be reduced. 객체 정보 추출 장치 및 방법이 개시된다. 일 실시예에 따른 객체 정보 추출 장치는, 원본 이미지를 포함하는 패딩(padding) 이미지를 생성하는 패딩 이미지 생성부, 상기 패딩 이미지에 대한 복수의 부분 이미지를 획득하는 부분 이미지 획득부, 객체 분류 모델을 이용하여 상기 복수의 부분 이미지 각각에 대한 객체 분류 결과를 획득하는 객체 분류 결과 획득부, 상기 패딩 이미지와 동일한 크기를 가지고, 상기 객체 분류 결과에 기초한 컨피던스 값을 포함하는 컨피던스 맵(confidence map)을 생성하는 컨피던스 맵 생성부 및 상기 컨피던스 맵에 기초하여 상기 패딩 이미지 내 객체 정보를 획득하는 객체 정보 획득부를 포함한다.
Bibliography:Application Number: KR20180130146