Open AccessReal-Time Object Detection based on Divergent Activation
Author(s) -
Hui Liang,
Yueyuan Wang,
Yi Ma,
Kewen Liu
Publication year - 2020
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/914/1/012022
Subject(s) - pascal (unit) , discriminative model , computer science , artificial intelligence , pattern recognition (psychology) , computer vision , object detection , programming language
Object detection is a research hotspot in the field of computer vision and has significant research value in the fields e.g. human-computer interaction, video surveillance, automatic driving, face recognition, medical imaging and etc. This paper proposes a novel real-time object detection algorithm based on YOLOv2 and divergent activation which aims to extract semantically complementary and discriminative features. The introduced divergent activation mechanism can be divided into differential divergent activation and hierarchical divergent activation that obtains the spatial complementary features by fusing different spatial distribution information of objects at the same level and activates a complete range of objects by fusing complementary semantics of objects from different levels, respectively. Experimental results on PASCAL VOC2007 and PASCAL VOC2012 show that our proposed algorithm reached 81.0% on mean average precision(mAP), indicating the effectiveness of the proposed algorithm.