arrow
返回

Task-decoupled interactive embedding network for object detection

delete2024-01-05
delete0
PRE
AI
M
Mai Liu
J
Jichao Jiao *
李
李宁 (Ning Li)
庞敏 封面图
庞敏 (Min Pang)
DOI:10.1007/s10994-023-06430-wdelete
delete原文链接
delete原文求助
delete分享
delete收藏
摘要

摘要

En 中文
Traditional object detection methods rely on manually annotated data, which can be costly and time-consuming, particularly for objects with low occurrence frequency or those that are neglected in existing datasets. When we need to generalize the model from the training datasets to the target datasets, false positive detection will appear with limited annotations in some categories and the model performance will decrease for unseen categories. In this paper, we found that the problems are related to the model's overfitting to foreground objects during the training stage and the inadequate robustness of feature representations. In order to effectively improve generalization of deep learning network, we propose a task-decoupled interactive embedding network. We decouple the sub-tasks in the detection pipeline with parallel convolution branches, with gradient propagation independently and anchor boxes generation from coarse to fine. And we introduce an embedding-interactive self-supervised decoder into the detector, so that the weaker object representations can be enhanced, and the representations of the same object can be closely aggregated, providing multi-scale semantic information for detection. Our method achieves great results on two visual tasks: few-shot object detection and open world object detection. It can effectively improve generalization on novel classes without hurting the detection of base classes and have good generalization ability for unknown categories detection. Our code is available at: https://github.com/hommelibrelm/DINet.
Keyword:
Object detection
Transformer
Task-decoupled network
Interactive embedding network

期刊

Machine Learning 封面图
Machine Learning
IF:
2.9
论文数:
2.7K
被引数:
3.4W

机构

B
beijing university of posts & telecommunications
学者数:
1.4W
论文数: 1.2W
被引数: 9
引用论文

引用论文

Psychosocial impact of illness intrusiveness moderated by self-concept and age in end-stage renal disease.
err1997-01-01
err0
PREAI
errGerald M. Devins; Heather Beanlands; Henry Mandin; Leendert C. Paul
err分享
err收藏
A glove-based system for studying hand-object manipulation via joint pose and force sensing
err2017-09-01
err0
PREAI
errHangxin Liu; Xu Xie; Matt Millar; Mark Edmonds; Feng Gao; Yixin Zhu; Veronica J. Santos; Brandon Rothrock; Song-Chun Zhu
err分享
err收藏
Automated evaluation of autoantibodies on human epithelial-2 cells as an approach to standardize cell-based immunofluorescence tests
err2010-03-09
err0
errOAAI
errKarl Egerer; Dirk Roggenbuck; Rico Hiemann; Max-Georg Weyer; Thomas Büttner; Boris Radau; Rosemarie Krause; Barbara Lehmann; Eugen Feist; Gerd-Rüdiger Burmester
err分享
err收藏
Structural Neural Substrates of Reading the Mind in the Eyes在眼睛中阅读心灵的结构神经基质
err2016-04-11
err0
errOAAI
errWataru Sato; Takanori Kochiyama; Shota Uono; Reiko Sawada; Yasutaka Kubota; Sayaka Yoshimura; Motomi Toichi
err分享
err收藏
Chronic exposure to glibenclamide impairs insulin secretion in isolated rat pancreatic islets
err2014-03-27
err0
PREAI
errDamiano Gullo; A. M. Rabuazzo; M. Vetri; C. Gatta; C. Vinci; M. Buscema; R. Vigneri; F. Purrello
err分享
err收藏
The PASCAL Visual Object Classes Challenge: A RetrospectivePASCAL视觉对象课程挑战: 回顾性
err2014-06-25
err5.0K
PREAI
errEveringham, Mark; Eslami, S. M. Ali; Van Gool, Luc; Williams, Christopher K. I.; Winn, John; Zisserman, Andrew
err分享
err收藏
Post-Genomic Update on a Classical Candidate Gene for Coronary Artery Disease: ESR1
err2011-12-01
err0
errOAAI
errGavin Lucas; Carla Lluís-Ganella; Isaac Subirana; Mariano Sentí; Christina Willenborg; Muntaser D. Musameh; Stephen M. Schwartz; Christopher J. O'Donnell; Olle Melander; Veikko Salomaa; Roberto Elosua
err分享
err收藏
学者 查看更多内容