DID: Disentangling-Imprinting-Distilling for Continuous Low-Shot Detection

doi:10.1109/TIP.2020.3006397

CORC > 自动化研究所 > 中国科学院自动化研究所 > 智能制造技术与系统研究中心

	DID: Disentangling-Imprinting-Distilling for Continuous Low-Shot Detection
	Chen, Xianyu 1,2; Wang, Yali 1,2; Liu, Jianzhuang 3; Qiao, Yu 1,2
刊名	IEEE TRANSACTIONS ON IMAGE PROCESSING
	2020
卷号	29 页码:7765-7778
关键词	Object detection low-shot learning continuous learning deep learning transfer learning
ISSN号	1057-7149
DOI	10.1109/TIP.2020.3006397
通讯作者	Qiao, Yu(yu.qiao@siat.ac.cn)
英文摘要	Practical applications often face a challenging continuous low-shot detection scenario, where a target detection task only has a few annotated training images, and a number of such new tasks come in sequence. To address this challenge, we propose a generic detection scheme via Disentangling-Imprinting-Distilling (DID). DID can leverage delicate transfer insights into the main development flow of deep learning, i.e., architecture design (Disentangling), model initialization (Imprinting), and training methodology (Distilling). This allows DID to be a simple but effective solution for continuous low-shot detection. In addition, DID can integrate the supervision from different detection tasks into a progressive learning procedure. As a result, one can efficiently adapt the previous detector for a new low-shot task, while maintaining the learned detection knowledge in the history. Finally, we evaluate our DID on a number of challenging settings in continuous/incremental low-shot detection. All the results demonstrate that our DID outperforms the recent state-of-the-art approaches. The code and models are available at https://github.com/chenxy99/DID.
资助项目	Science and Technology Service Network Initiative of Chinese Academy of Sciences[KFJ-STS-QYZX-092] ; National Natural Science Foundation of China[61876176] ; National Natural Science Foundation of China[U1713208] ; Shenzhen Basic Research Program[JCYJ20170818164704758] ; Shenzhen Basic Research Program[CXB201104220032A] ; Shenzhen Institute of Artificial Intelligence and Robotics for Society
WOS研究方向	Computer Science ; Engineering
语种	英语
出版者	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
WOS记录号	WOS:000549387700006
资助机构	Science and Technology Service Network Initiative of Chinese Academy of Sciences ; National Natural Science Foundation of China ; Shenzhen Basic Research Program ; Shenzhen Institute of Artificial Intelligence and Robotics for Society
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/40177]
专题	自动化研究所_智能制造技术与系统研究中心
通讯作者	Qiao, Yu
作者单位	1.Chinese Acad Sci, Shenzhen Inst Adv Technol, Shenzhen Key Lab Comp Vis & Pattern Recognit, Shenzhen 518055, Peoples R China 2.Shenzhen Inst Artificial Intelligence & Robot Soc, SIAT Branch, Shenzhen 518055, Peoples R China 3.Huawei Technol Co Ltd, Noahs Ark Lab, Shenzhen 518129, Peoples R China
推荐引用方式 GB/T 7714	Chen, Xianyu,Wang, Yali,Liu, Jianzhuang,et al. DID: Disentangling-Imprinting-Distilling for Continuous Low-Shot Detection[J]. IEEE TRANSACTIONS ON IMAGE PROCESSING,2020,29:7765-7778.
APA	Chen, Xianyu,Wang, Yali,Liu, Jianzhuang,&Qiao, Yu.(2020).DID: Disentangling-Imprinting-Distilling for Continuous Low-Shot Detection.IEEE TRANSACTIONS ON IMAGE PROCESSING,29,7765-7778.
MLA	Chen, Xianyu,et al."DID: Disentangling-Imprinting-Distilling for Continuous Low-Shot Detection".IEEE TRANSACTIONS ON IMAGE PROCESSING 29(2020):7765-7778.