中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
首页 机构 成果 学者
  
  1. CAS IR Grid
  2. 自动化研究所
  3. 中国科学院自动化研究所
  4. 智能感知与计算研究中心
Toward few-shot domain adaptation with perturbation-invariant representation and transferable prototypes

文献类型：期刊论文

作者Fan, Junsong1,3; Wang, Yuxi1,2,3; Guan, He1,3; Song, Chunfeng1,3; Zhang, Zhaoxiang1,2,3
刊名FRONTIERS OF COMPUTER SCIENCE
出版日期2022-06-01
卷号16期号:3页码:11
关键词domain adaptation semantic segmentation
ISSN号2095-2228
DOI10.1007/s11704-022-2015-7
通讯作者Zhang, Zhaoxiang(zhaoxiang.zhang@ia.ac.cn)
英文摘要Domain adaptation (DA) for semantic segmentation aims to reduce the annotation burden for the dense pixel-level prediction task. It focuses on tackling the domain gap problem and manages to transfer knowledge learned from abundant source data to new target scenes. Although recent works have achieved rapid progress in this field, they still underperform fully supervised models with a large margin due to the absence of any available hints in the target domain. Considering that few-shot labels are cheap to obtain in practical applications, we attempt to leverage them to mitigate the performance gap between DA and fully supervised methods. The key to this problem is to leverage the few-shot labels to learn robust domain-invariant predictions effectively. To this end, we first design a data perturbation strategy to enhance the robustness of the representations. Furthermore, a transferable prototype module is proposed to bridge the domain gap based on the source data and few-shot targets. By means of these proposed methods, our approach can perform on par with the fully supervised models to some extent. We conduct extensive experiments to demonstrate the effectiveness of the proposed methods and report the state-of-the-art performance on two popular DA tasks, i.e., from GTA5 to Cityscapes and SYNTHIA to Cityscapes.
资助项目National Key R&D Program of China[2019QY1604] ; Major Project for New Generation of AI[2018AAA0100400] ; National Youth Talent Support Program ; National Natural Science Foundation of China[U21B2042] ; National Natural Science Foundation of China[62006231] ; National Natural Science Foundation of China[62072457]
WOS研究方向Computer Science
语种英语
WOS记录号WOS:000789054200001
出版者HIGHER EDUCATION PRESS
资助机构National Key R&D Program of China ; Major Project for New Generation of AI ; National Youth Talent Support Program ; National Natural Science Foundation of China
源URL[http://ir.ia.ac.cn/handle/173211/48419]  
专题自动化研究所_智能感知与计算研究中心
通讯作者Zhang, Zhaoxiang
作者单位1.Univ Chinese Acad Sci, Beijing 100049, Peoples R China
2.HKISI CAS, Ctr Artificial Intelligence & Robot, Hong Kong 999077, Peoples R China
3.Chinese Acad Sci, Ctr Res Intelligent Percept & Comp, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China
推荐引用方式
GB/T 7714		 Fan, Junsong,Wang, Yuxi,Guan, He,et al. Toward few-shot domain adaptation with perturbation-invariant representation and transferable prototypes[J]. FRONTIERS OF COMPUTER SCIENCE,2022,16(3):11.
APA Fan, Junsong,Wang, Yuxi,Guan, He,Song, Chunfeng,&Zhang, Zhaoxiang.(2022).Toward few-shot domain adaptation with perturbation-invariant representation and transferable prototypes.FRONTIERS OF COMPUTER SCIENCE,16(3),11.
MLA Fan, Junsong,et al."Toward few-shot domain adaptation with perturbation-invariant representation and transferable prototypes".FRONTIERS OF COMPUTER SCIENCE 16.3(2022):11.

入库方式： OAI收割

来源：自动化研究所

浏览0
下载0
收藏0
其他版本

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。