文献类型：期刊文献

英文题名：SWIPENET: Object detection in noisy underwater scenes

作者：Chen, Long[1];Zhou, Feixiang[1];Wang, Shengke[2];Dong, Junyu[2];Li, Ning[3];Ma, Haiping[4];Wang, Xin[5];Zhou, Huiyu[1]

机构：[1]Univ Leicester, Sch Comp & Math Sci, Leicester, England;[2]Ocean Univ China, Dept informat Sci & Engn, Qingdao, Peoples R China;[3]Nanjing Univ Aeronaut & Astronaut, Coll Elect & Informat Engn, Nanjing, Peoples R China;[4]Shaoxing Univ, Dept Elect Engn, Shaoxing, Peoples R China;[5]Hohai Univ, Coll Comp & Informat, Nanjing, Peoples R China

年份：2022

卷号：132

外文期刊名：PATTERN RECOGNITION

收录：SCI-EXPANDED(收录号：WOS:000860613700007)、、EI(收录号：20223112474281)、Scopus(收录号：2-s2.0-85135150430)、WOS

基金：Thanks for National Natural Science Foundation of China and Dalian Municipal Peoples Government providing the underwater object detection datasets for research purposes. Haiping Ma is sup-ported by Zhejiang Provincial Natural Science Foundation of China under Grant No. LY19F030011.

语种：英文

外文关键词：Underwater object detection; Curriculum Multi -Class Adaboost; Sample -weighted detection loss; Noisy data

外文摘要：Deep learning based object detection methods have achieved promising performance in controlled envi-ronments. However, these methods lack sufficient capabilities to handle underwater object detection due to these challenges: (1) images in the underwater datasets and real applications are blurry whilst accom-panying severe noise that confuses the detectors and (2) objects in real applications are usually small. In this paper, we propose a Sample-WeIghted hyPEr Network (SWIPENET), and a novel training paradigm named Curriculum Multi-Class Adaboost (CMA), to address these two problems at the same time. Firstly, the backbone of SWIPENET produces multiple high resolution and semantic-rich Hyper Feature Maps, which significantly improve small object detection. Secondly, inspired by the human education process that drives the learning from easy to hard concepts, we propose the noise-robust CMA training paradigm that learns the clean data first and then move on to learns the diverse noisy data. Experiments on four underwater object detection datasets show that the proposed SWIPENET+CMA framework achieves better or competitive accuracy in object detection against several state-of-the-art approaches.(c) 2022 Elsevier Ltd. All rights reserved.