METHOD AND SYSTEM FOR GENERATING TRAINING DATA TO TRAIN CLASSIFIERS WITH LOCALIZABLE FEATURES

Disclosed are a training data creation method for training a classifier with regional characteristics and a system therefor. The training data creation method comprises the following steps of: removing some areas from one sample image of two sample images, mixing the two sample images by using a met...

Full description

Saved in:
Bibliographic Details
Main Authors CHUN SANGHYUK, YUN SANGDOO, HAN DONGYOON, YOO YOUNGJOON
Format Patent
LanguageEnglish
Korean
Published 28.10.2020
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:Disclosed are a training data creation method for training a classifier with regional characteristics and a system therefor. The training data creation method comprises the following steps of: removing some areas from one sample image of two sample images, mixing the two sample images by using a method of replacing the removed areas to a patch of the other sample image to create a new image; and training a convolutional neural network (CNN) model by using the created image as training data. Therefore, classifier performance and regional characteristic recognition performance can be simultaneously improved by creating a new training image by using a method of cutting and pasting images. 지역적 특징을 가지는 분류기 학습을 위한 학습 데이터 생성 방법 및 그 시스템이 개시된다. 학습 데이터 생성 방법은, 두 개의 샘플 이미지 중 하나의 샘플 이미지에서 일부 영역을 제거한 후 제거된 영역을 다른 하나의 샘플 이미지의 패치(patch)로 대체하는 방식으로 상기 두 개의 샘플 이미지를 혼합하여(mix) 새로운 이미지를 생성하는 단계; 및 상기 생성된 이미지를 학습 데이터로 사용하여 CNN(convolutional neural network) 모델을 학습하는 단계를 포함한다.
Bibliography:Application Number: KR20190054763