PDNet: Toward Better One-Stage Object Detection With Prediction Decoupling

doi:10.1109/TIP.2022.3193223

CORC > 自动化研究所 > 中国科学院自动化研究所 > 模式识别国家重点实验室 > 视频内容安全团队

	PDNet: Toward Better One-Stage Object Detection With Prediction Decoupling
	Yang, Li 4,5; Xu, Yan 3; Wang, Shaoru 4,5; Yuan, Chunfeng 5; Zhang, Ziqi 4,5; Li, Bing 2,5; Hu, Weiming 1,4,5
刊名	IEEE TRANSACTIONS ON IMAGE PROCESSING
	2022
卷号	31 页码:5121-5133
关键词	Object detection prediction decoupling convolutional neural network
ISSN号	1057-7149
DOI	10.1109/TIP.2022.3193223
通讯作者	Yuan, Chunfeng(cfyuan@nlpr.ia.ac.cn)
英文摘要	Recent one-stage object detectors follow a per-pixel prediction approach that predicts both the object category scores and boundary positions from every single grid location. However, the most suitable positions for inferring different targets, i.e., the object category and boundaries, are generally different. Predicting all these targets from the same grid location thus may lead to sub-optimal results. In this paper, we analyze the suitable inference positions for object category and boundaries, and propose a prediction-target-decoupled detector named PDNet to establish a more flexible detection paradigm. Our PDNet with the prediction decoupling mechanism encodes different targets separately in different locations. A learnable prediction collection module is devised with two sets of dynamic points, i.e., dynamic boundary points and semantic points, to collect and aggregate the predictions from the favorable regions for localization and classification. We adopt a two-step strategy to learn these dynamic point positions, where the prior positions are estimated for different targets first, and the network further predicts residual offsets to the positions with better perceptions of the object properties. Extensive experiments on the MS COCO benchmark demonstrate the effectiveness and efficiency of our method. With a single ResNeXt-64x4d-101-DCN as the backbone, our detector achieves 50.1 AP with single-scale testing, which outperforms the state-of-the-art methods by an appreciable margin under the same experimental settings. Moreover, our detector is highly efficient as a one-stage framework. Our code is public at https://github.com/yangli18/PDNet.
资助项目	National Key Research and Development Program of China[2020AAA0106800] ; Beijing Natural Science Foundation[JQ21017] ; Beijing Natural Science Foundation[4224091] ; National Natural Science Foundation of China[61972397] ; National Natural Science Foundation of China[62036011] ; National Natural Science Foundation of China[62192782] ; National Natural Science Foundation of China[61721004] ; National Natural Science Foundation of China[61906192] ; Key Research Program of Frontier Sciences, CAS[QYZDJ-SSW-JSC040] ; China Postdoctoral Science Foundation[2021M693402]
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:000835774000011
资助机构	National Key Research and Development Program of China ; Beijing Natural Science Foundation ; National Natural Science Foundation of China ; Key Research Program of Frontier Sciences, CAS ; China Postdoctoral Science Foundation
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/49821]
专题	自动化研究所_模式识别国家重点实验室_视频内容安全团队
通讯作者	Yuan, Chunfeng
作者单位	1.CAS Ctr Excellence Brain Sci & Intelligence Techn, Shanghai 200031, Peoples R China 2.PeopleAI Inc, Beijing 100190, Peoples R China 3.Chinese Univ Hong Kong, Dept Elect Engn, Hong Kong, Peoples R China 4.Univ Chinese Acad Sci, Sch Artificial Intelligence, Beijing 100190, Peoples R China 5.Chinese Acad Sci, Natl Lab Pattern Recognit, Inst Automat, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Yang, Li,Xu, Yan,Wang, Shaoru,et al. PDNet: Toward Better One-Stage Object Detection With Prediction Decoupling[J]. IEEE TRANSACTIONS ON IMAGE PROCESSING,2022,31:5121-5133.
APA	Yang, Li.,Xu, Yan.,Wang, Shaoru.,Yuan, Chunfeng.,Zhang, Ziqi.,...&Hu, Weiming.(2022).PDNet: Toward Better One-Stage Object Detection With Prediction Decoupling.IEEE TRANSACTIONS ON IMAGE PROCESSING,31,5121-5133.
MLA	Yang, Li,et al."PDNet: Toward Better One-Stage Object Detection With Prediction Decoupling".IEEE TRANSACTIONS ON IMAGE PROCESSING 31(2022):5121-5133.