Progressive Object Transfer Detection
文献类型：期刊论文
| 作者 | Chen, Hao2,6 ; Wang, Yali3,4; Wang, Guoyou5; Bai, Xiang1; Qiao, Yu3,4
|
| 刊名 | IEEE TRANSACTIONS ON IMAGE PROCESSING
 |
| 出版日期 | 2020 |
| 卷号 | 29页码:986-1000 |
| 关键词 | Detectors Object detection Proposals Task analysis Benchmark testing Deep learning Labeling Object detection deep learning transfer learning weakly semi-supervised detection low-shot learning |
| ISSN号 | 1057-7149 |
| DOI | 10.1109/TIP.2019.2938680 |
| 通讯作者 | Qiao, Yu(yu.qiao@siat.ac.cn) |
| 英文摘要 | Recent development of object detection mainly depends on deep learning with large-scale benchmarks. However, collecting such fully-annotated data is often difficult or expensive for real-world applications, which restricts the power of deep neural networks in practice. Alternatively, humans can detect new objects with little annotation burden, since humans often use the prior knowledge to identify new objects with few elaborately-annotated examples, and subsequently generalize this capacity by exploiting objects from wild images. Inspired by this procedure of learning to detect, we propose a novel Progressive Object Transfer Detection (POTD) framework. Specifically, we make three main contributions in this paper. First, POTD can leverage various object supervision of different domains effectively into a progressive detection procedure. Via such human-like learning, one can boost a target detection task with few annotations. Second, POTD consists of two delicate transfer stages, i.e., Low-Shot Transfer Detection (LSTD), and Weakly-Supervised Transfer Detection (WSTD). In LSTD, we distill the implicit object knowledge of source detector to enhance target detector with few annotations. It can effectively warm up WSTD later on. In WSTD, we design a recurrent object labelling mechanism for learning to annotate weakly-labeled images. More importantly, we exploit the reliable object supervision from LSTD, which can further enhance the robustness of target detector in the WSTD stage. Finally, we perform extensive experiments on a number of challenging detection benchmarks with different settings. The results demonstrate that, our POTD outperforms the recent state-of-the-art approaches. The codes and models are available at |
| 资助项目 | National Key Research and Development Program of China[2016YFC1400704] ; National Natural Science Foundation of China[61876176] ; National Natural Science Foundation of China[U1613211] ; National Natural Science Foundation of China[U1713208] ; Shenzhen Basic Research Program[JCYJ20170818164704758] ; Joint Lab of CAS-HK |
| WOS研究方向 | Computer Science ; Engineering |
| 语种 | 英语 |
| WOS记录号 | WOS:000498872600001 |
| 出版者 | IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC |
| 资助机构 | National Key Research and Development Program of China ; National Natural Science Foundation of China ; Shenzhen Basic Research Program ; Joint Lab of CAS-HK |
| 源URL | [http://ir.ia.ac.cn/handle/173211/29331]  |
| 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.Huazhong Univ Sci & Technol, Sch Elect Informat & Commun, Wuhan 430074, Peoples R China 2.Univ Maryland, Dept Comp Sci, College Pk, MD 20740 USA 3.Chinese Acad Sci, Shenzhen Inst Adv Technol, Shenzhen Key Lab CVPR, Shenzhen 518055, Guangdong, Peoples R China 4.Chinese Acad Sci, Shenzhen Inst Adv Technol, SIAT Sensetime Joint Lab, Shenzhen 518055, Guangdong, Peoples R China 5.Huazhong Univ Sci & Technol, Sch Artificial Intelligence & Automat, Wuhan 430074, Peoples R China 6.Chinese Acad Sci, Shenzhen Inst Adv Technol, Shenzhen 518055, Guangdong, Peoples R China |
| 推荐引用方式 GB/T 7714 | Chen, Hao,Wang, Yali,Wang, Guoyou,et al. Progressive Object Transfer Detection[J]. IEEE TRANSACTIONS ON IMAGE PROCESSING,2020,29:986-1000. |
| APA | Chen, Hao,Wang, Yali,Wang, Guoyou,Bai, Xiang,&Qiao, Yu.(2020).Progressive Object Transfer Detection.IEEE TRANSACTIONS ON IMAGE PROCESSING,29,986-1000. |
| MLA | Chen, Hao,et al."Progressive Object Transfer Detection".IEEE TRANSACTIONS ON IMAGE PROCESSING 29(2020):986-1000. |
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。