Integral Feature를 적용한 효율적인 Convolutional Neural Network 기반 객체 위치 탐색 기법

Abstract

Convolutional Neural Network(CNN) is the state-of-the-art object recognition method in computer vision research area. But CNN has a very complicated structure and high cost in image localization. We present an efficient method for classification and localization of objects in convolutional neural network. The proposed method is faster than the conventional method in image searching, while achieving the capability of CNN.