跳到主要导航 跳到搜索 跳到主要内容

Effective Feature Enhancement and Model Ensemble Strategies in Tiny Object Detection

  • Yuan Feng
  • , Xiaodi Wang
  • , Ying Xin
  • , Bin Zhang
  • , Jingwei Liu
  • , Mingyuan Mao
  • , Sheng Xu
  • , Baochang Zhang
  • , Shumin Han*
  • *此作品的通讯作者
  • Baidu Inc
  • Chinese Academy of Sciences
  • Beihang University

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

We introduce a novel tiny-object detection network that achieves better accuracy than existing detectors on TinyPerson dataset. It is an end-to-end detection framework developed on PaddlePaddle. A suit of strategies are developed to improve the detectors performance including: 1) data augmentation based on scale-match that aligns the object scales between the existing large-scale dataset and TinyPerson; 2) comprehensive training methods to further improve detection performance by a large margin; 3) model refinement based on the enhanced PAFPN module to fully utilize semantic information; 4) a hierarchical coarse-to-fine ensemble strategy to improve detection performance based on a well-designed model pond.

源语言英语
主期刊名Computer Vision – ECCV 2020 Workshops, Proceedings
编辑Adrien Bartoli, Andrea Fusiello
出版商Springer Science and Business Media Deutschland GmbH
324-330
页数7
ISBN(印刷版)9783030682378
DOI
出版状态已出版 - 2020
活动Workshops held at the 16th European Conference on Computer Vision, ECCV 2020 - Glasgow, 英国
期限: 23 8月 202028 8月 2020

出版系列

姓名Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
12539 LNCS
ISSN(印刷版)0302-9743
ISSN(电子版)1611-3349

会议

会议Workshops held at the 16th European Conference on Computer Vision, ECCV 2020
国家/地区英国
Glasgow
时期23/08/2028/08/20

指纹

探究 'Effective Feature Enhancement and Model Ensemble Strategies in Tiny Object Detection' 的科研主题。它们共同构成独一无二的指纹。

引用此