深度学习应用于遮挡目标检测算法综述

doi:10.3778/j.issn.1673-9418.2112035

计算机科学与探索 ›› 2022, Vol. 16 ›› Issue (6): 1243-1259.DOI: 10.3778/j.issn.1673-9418.2112035

深度学习应用于遮挡目标检测算法综述

孙方伟¹, 李承阳¹^,², 谢永强¹^,⁺(), 李忠博¹, 杨才东¹, 齐锦¹

1. 军事科学院系统工程研究院,北京 100141
2. 北京大学信息科学技术学院,北京 100871

收稿日期:2021-12-09 修回日期:2022-03-15 出版日期:2022-06-01 发布日期:2022-06-20
通讯作者: + E-mail: xyq_ams@outlook.com
作者简介:孙方伟（1996—）,男,山东青岛人,硕士研究生,主要研究方向为目标检测、目标跟踪、语义分割。
李承阳（1995—）,男,辽宁鞍山人,博士研究生,主要研究方向为目标检测、数据挖掘、机器学习。
谢永强（1972—）,男,北京人,博士,研究员,主要研究方向为机器学习、多媒体技术、智能系统架构等。
李忠博（1983—）,男,北京人,博士,高级工程师,主要研究方向为多媒体技术、机器学习、云视频等。
杨才东（1996—）,男,贵州六盘水人,硕士研究生,主要研究方向为超分辨率重建。
齐锦（1971—）,女,北京人,硕士,高级工程师,主要研究方向为多媒体技术、智能系统架构、云计算等。

Review of Deep Learning Applied to Occluded Object Detection

SUN Fangwei¹, LI Chengyang¹^,², XIE Yongqiang¹^,⁺(), LI Zhongbo¹, YANG Caidong¹, QI Jin¹

1. Academy of Systems Engineering, Academy of Military Sciences, Beijing 100141, China
2. School of Electronics Engineering and Computer Science, Peking University, Beijing 100871, China

Received:2021-12-09 Revised:2022-03-15 Online:2022-06-01 Published:2022-06-20
About author:SUN Fangwei, born in 1996, M.S. candidate. His research interests include object detection, object tracking and semantic segmentation.
LI Chengyang, born in 1995, Ph.D. candidate. His research interests include object detection, data mining and machine learning.
XIE Yongqiang, born in 1972, Ph.D., researcher. His research interests include machine learning, multimedia technology, intelligent system architecture, etc.
LI Zhongbo, born in 1983, Ph.D., senior engineer. His research interests include multimedia technology, machine learning, cloud video, etc.
YANG Caidong, born in 1996, M.S. candidate. His research interest is super-resolution reconstruction.
QI Jin, born in 1971, M.S., senior engineer. Her research interests include multimedia technology, intelligent system architecture, cloud computing, etc.

摘要/Abstract

摘要：

遮挡目标检测长期以来是计算机视觉中的一个难点和研究热点。目前的深度学习基于卷积神经网络,将目标检测任务作为分类任务和回归任务来处理。当目标被遮挡时,遮挡物会混淆目标之间的特征,使得深度网络不能很好地识别和推理,降低检测器在理想场景下的性能。考虑到遮挡在现实中的普遍性,对遮挡目标的有效检测具有重要研究价值。为了进一步促进遮挡目标检测的发展,对基于深度学习的遮挡目标检测算法进行了全面总结,并对已有的遮挡检测算法进行归类、分析、比较。在对目标检测进行简单概述基础上,首先,对遮挡目标检测的相关背景、研究的难点以及遮挡数据集进行了介绍;然后,对遮挡检测优化算法主要按照目标结构、损失函数、非极大值抑制以及部分语义四方面进行归纳分析,在对各种算法之间的联系以及发展脉络进行阐述后,对各种算法性能进行了比较;最后,指出了遮挡目标检测仍面临的困难,并对遮挡目标检测未来的发展方向进行了展望。

关键词: 遮挡目标检测, 深度学习, 损失函数, 非极大值抑制, 部分语义

Abstract:

Occluded object detection has long been a difficulty and hot topic in the field of computer vision. Based on convolutional neural network, the deep learning takes the object detection task as a classification and regression task to handle, and obtains remarkable achievements. The mask confuses the features of object when the object is occluded, making the deep convolutional neural network cannot handle it well and reducing the performance of detector in ideal scenes. Considering the universality of occlusion in reality, the effective detection of occluded object has important research value. In order to further promote the development of occluded object detection, this paper makes a comprehensive summary of occluded object detection algorithms, and makes a reasonable classification and analysis. First of all, based on a simple overview of object detection, this paper introduces the relevant theoretic background, research difficulties and datasets about occluded object detection. After, this paper focuses on the algo-rithms to improve the performance of occluded object detection from the aspects of object structure, loss function, non-maximum suppression and semantic partial. This paper compares the performance of different detection algo-rithms after summarizing the relationship and development of various algorithms. Finally, this paper points out the difficulties of occluded object detection and looks forward to its future development directions.

Key words: occluded object detection, deep learning, loss function, non-maximum suppression, semantic partial

中图分类号:

TP391

孙方伟, 李承阳, 谢永强, 李忠博, 杨才东, 齐锦. 深度学习应用于遮挡目标检测算法综述[J]. 计算机科学与探索, 2022, 16(6): 1243-1259.

SUN Fangwei, LI Chengyang, XIE Yongqiang, LI Zhongbo, YANG Caidong, QI Jin. Review of Deep Learning Applied to Occluded Object Detection[J]. Journal of Frontiers of Computer Science and Technology, 2022, 16(6): 1243-1259.

图/表 14

表1 遮挡目标检测数据集

Table 1 Datasets of occlusion object detection

数据集	年份	图片数量	目标类别	每张图片目标个数	每张图片遮挡目标个数	各部分占比/%			使用场景
数据集	年份	图片数量	目标类别	每张图片目标个数	每张图片遮挡目标个数	训练	验证	测试	使用场景
ImageNet	2009	14 197 122	21 841	—	—	—	—	—	综合
PASCAL VOC	2007	9 963	20	2.47	—	25.0	25.0	50.0	综合
PASCAL VOC	2012	23 080	20	2.38	—	25.0	25.0	50.0	综合
MS-COCO	2014	328 000	80	9.34	0.015	50.0	25.0	25.0	综合
Open Images	2018	9 178 276	~6 000	8.00	—	98.2	1.8	—	综合
KITTI	2012	14 999	5	5.35	—	—	—	—	行人、车辆
Caltech数据集	2012	250 000	1	0.32	0.320	50.0	50.0	行人
VehicleOcclusion	2017	9 056	6	1.00	1.000	50.0	50.0	车辆
CityPersons	2017	5 050	1	6.47	0.320	~58.9	~9.9	~31.2	行人
CrowdHuman	2018	24 370	1	22.64	2.400	61.6	17.9	20.5	行人

图1 算法分类

Fig.1 Algorithm classification

图2 PORoI和遮挡处理单元结构图

Fig.2 Architecture of PORoI and occlusion process unit

图3 CoupleNet模型结构

Fig.3 Architecture of CoupleNet

图4 JointDet网络结构

Fig.4 Network structure of JointDet

图5 DA-RCNN结构图

Fig.5 Network structure of DA-RCNN

图6 多目标预测总体结构

Fig.6 Architecture of multi-target prediction

表2 损失函数结构

Table 2 Architecture of loss function

损失函数	函数结构	结构说明
Repulsion Loss	$L R e p = L A t t r + α ∙ L R e p G T + β ∙ L R e p B o x$	第一部分使得预测框尽可能靠近目标框第二部分使得预测框尽可能远离周围目标框第三部分使得预测框尽可能远离周围其他预测框
Aggregation Loss	$L a g g = L r e g + β ∙ L c o m$	第一部分为回归损失,使得建议框靠近目标框第二部分使得同一目标的建议框尽可能靠近
GIoU Loss	$L G I o U = 1 - G I o U$	使用交并比GIoU进行损失计算
Rep-GIoU Loss	$L R e p - G I o U = L G I o U + α ∙ L R e p$	GIoU Loss和Repulsion Loss的组合函数
NMS Loss	$L N M S = α ∙ L p u l l + β ∙ L p u s h$	第一部分用来抑制假阳性第二部分用来避免错误地删除假阴性

表2 损失函数结构

Table 2 Architecture of loss function

损失函数	函数结构	结构说明
Repulsion Loss	$L R e p = L A t t r + α ∙ L R e p G T + β ∙ L R e p B o x$	第一部分使得预测框尽可能靠近目标框第二部分使得预测框尽可能远离周围目标框第三部分使得预测框尽可能远离周围其他预测框
Aggregation Loss	$L a g g = L r e g + β ∙ L c o m$	第一部分为回归损失,使得建议框靠近目标框第二部分使得同一目标的建议框尽可能靠近
GIoU Loss	$L G I o U = 1 - G I o U$	使用交并比GIoU进行损失计算
Rep-GIoU Loss	$L R e p - G I o U = L G I o U + α ∙ L R e p$	GIoU Loss和Repulsion Loss的组合函数
NMS Loss	$L N M S = α ∙ L p u l l + β ∙ L p u s h$	第一部分用来抑制假阳性第二部分用来避免错误地删除假阴性

图7 NMS算法的结构

Fig.7 Architecture of NMS

图8 SENet总体结构

Fig.8 Overall architecture of SENet

图9 DeepVoting整体框架

Fig.9 Overall framework of DeepVoting

图10 TDAPNet结构图

Fig.10 Overall architecture of TDAPNet

表3 不同类型改进方案的比较

Table 3 Comparison of different types of programmes

算法类型		优势	局限性
基于数据增强		实现简单,可作为另外两种方式的数据预处理手段	不符合现实情形,可能导致必要特征的缺失
基于整体特征	基于目标结构	利用可见部分,有效降低遮挡物的干扰,特征信息利用率高,分类能力强	部件检测器消耗计算资源,数据集要求高,鲁棒性低
	基于损失函数	可解释性,代价小	针对遮挡情形的设计难度较高
	基于非极大值抑制	适用范围广	不同场景的阈值设定具有差异化
基于部分语义	鲁棒性高,对遮挡的适应性强	分类能力较弱,空间语义信息的获取难度较高

表4 遮挡检测算法的性能

Table 4 Performance of occlusion detection algorithms

遮挡检测算法	提出时间	数据集	AP/%	MR^-2/%
DeepParts	2015	KITTI	58.7（部分遮挡）	—
DeepParts	2015	Caltech	—	12.9
OR-CNN	2018	CityPersons	—	5.9（轻度遮挡） 11.0（一般遮挡） 13.7（部分遮挡） 51.3（严重遮挡）
OR-CNN	2018	Caltech	—	4.1
CoupleNet	2017	PASCAL VOC	82.7	—
CoupleNet	2017	MS-COCO	34.4	—
文献[52]	2021	CityPersons	—	12.4（一般遮挡） 38.3（部分遮挡） 49.8（严重遮挡）
文献[52]	2021	Caltech	—	4.7（一般遮挡） 40.7（部分遮挡） 34.6（严重遮挡）
JointDet	2019	Caltech	—	2.9
		CrowdHuman	—	46.5
		CityPersons	—	10.2
DA-RCNN	2019	CrowdHuman	—	51.8
CrowdDet	2020	CrowdHuman	90.7	41.4
		CityPersons	96.1	10.7
		MS-COCO	38.5	—
MFRN	2021	CrowdHuman	90.9	40.2
MFRN	2021	CityPersons	96.2	10.6
Repulsion Loss	2017	CityPersons	—	13.2
		Caltech	—	4.0
		CrowdHuman	—	54.6
NMS Loss	2021	CityPersons	—	10.1
NMS Loss	2021	Caltech	—	5.9
Rep-GIoU Loss	2022	PASCAL VOC	82.9	—
FPN+NMS	2017	CrowdHuman	88.1	42.9
FPN+NMS	2017	CityPersons	95.2	11.8
FPN+Soft-NMS	2017	CrowdHuman	88.2	42.9
		CityPersons	95.3	11.8
		MS-COCO	38.0	—
GossipNet	2017	CrowdHuman	80.4	49.4
		PETS	81.4	—
		MS-COCO	66.6	—
Softer-NMS	2018	MS-COCO	40.4	—
Adaptive-NMS	2019	CrowdHuman	84.7	49.7
Adaptive-NMS	2019	CityPersons	—	10.8
R²NMS	2020	CrowdHuman	89.3	43.4
文献[72]	2021	CityPersons	—	9.3
文献[72]	2021	Caltech	—	6.8
文献[74]	2021	CityPersons	—	26.1
		Caltech	—	4.5
		CrowHuman	—	45.1
遮挡检测算法	提出时间	数据集	AP/%	MR^-2/%
DeepVoting	2018	VehicleSemanticPart	—	74.0（无遮挡） 58.0（20%~40%） 46.9（40%~60%） 35.2（60%~80%）
CompositionalNets^[77]	2020	PASCAL3D+	89.5	—
CompositionalNets^[77]	2020	MNIST	69.4	—
TDAPNet	2019	PASCAL3D+	92.8	—
TDAPNet	2019	MNIST	69.3	—
Kortylewski等^[79]	2020	PASCAL3D+	95.4	—
Kortylewski等^[79]	2020	Occluded-MS-COCO	94.4	—
Kortylewski等^[80]	2021	PASCAL3D+	84.1	—
Kortylewski等^[80]	2021	Occluded-MS-COCO	95.0	—
Wang等^[81]	2020	OccludedVehiclesDetection	81.4	—
Wang等^[81]	2020	OccludedCOCO	91.8（0~20%） 83.6（20%~40%） 77.8（40%~60%） 65.4（60%~80%） 59.6（80%~100%）	—

表4 遮挡检测算法的性能

Table 4 Performance of occlusion detection algorithms

遮挡检测算法	提出时间	数据集	AP/%	MR^-2/%
DeepParts	2015	KITTI	58.7（部分遮挡）	—
DeepParts	2015	Caltech	—	12.9
OR-CNN	2018	CityPersons	—	5.9（轻度遮挡） 11.0（一般遮挡） 13.7（部分遮挡） 51.3（严重遮挡）
OR-CNN	2018	Caltech	—	4.1
CoupleNet	2017	PASCAL VOC	82.7	—
CoupleNet	2017	MS-COCO	34.4	—
文献[52]	2021	CityPersons	—	12.4（一般遮挡） 38.3（部分遮挡） 49.8（严重遮挡）
文献[52]	2021	Caltech	—	4.7（一般遮挡） 40.7（部分遮挡） 34.6（严重遮挡）
JointDet	2019	Caltech	—	2.9
		CrowdHuman	—	46.5
		CityPersons	—	10.2
DA-RCNN	2019	CrowdHuman	—	51.8
CrowdDet	2020	CrowdHuman	90.7	41.4
		CityPersons	96.1	10.7
		MS-COCO	38.5	—
MFRN	2021	CrowdHuman	90.9	40.2
MFRN	2021	CityPersons	96.2	10.6
Repulsion Loss	2017	CityPersons	—	13.2
		Caltech	—	4.0
		CrowdHuman	—	54.6
NMS Loss	2021	CityPersons	—	10.1
NMS Loss	2021	Caltech	—	5.9
Rep-GIoU Loss	2022	PASCAL VOC	82.9	—
FPN+NMS	2017	CrowdHuman	88.1	42.9
FPN+NMS	2017	CityPersons	95.2	11.8
FPN+Soft-NMS	2017	CrowdHuman	88.2	42.9
		CityPersons	95.3	11.8
		MS-COCO	38.0	—
GossipNet	2017	CrowdHuman	80.4	49.4
		PETS	81.4	—
		MS-COCO	66.6	—
Softer-NMS	2018	MS-COCO	40.4	—
Adaptive-NMS	2019	CrowdHuman	84.7	49.7
Adaptive-NMS	2019	CityPersons	—	10.8
R²NMS	2020	CrowdHuman	89.3	43.4
文献[72]	2021	CityPersons	—	9.3
文献[72]	2021	Caltech	—	6.8
文献[74]	2021	CityPersons	—	26.1
		Caltech	—	4.5
		CrowHuman	—	45.1
遮挡检测算法	提出时间	数据集	AP/%	MR^-2/%
DeepVoting	2018	VehicleSemanticPart	—	74.0（无遮挡） 58.0（20%~40%） 46.9（40%~60%） 35.2（60%~80%）
CompositionalNets^[77]	2020	PASCAL3D+	89.5	—
CompositionalNets^[77]	2020	MNIST	69.4	—
TDAPNet	2019	PASCAL3D+	92.8	—
TDAPNet	2019	MNIST	69.3	—
Kortylewski等^[79]	2020	PASCAL3D+	95.4	—
Kortylewski等^[79]	2020	Occluded-MS-COCO	94.4	—
Kortylewski等^[80]	2021	PASCAL3D+	84.1	—
Kortylewski等^[80]	2021	Occluded-MS-COCO	95.0	—
Wang等^[81]	2020	OccludedVehiclesDetection	81.4	—
Wang等^[81]	2020	OccludedCOCO	91.8（0~20%） 83.6（20%~40%） 77.8（40%~60%） 65.4（60%~80%） 59.6（80%~100%）	—

参考文献 83

[1]	GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich fea-ture hierarchies for accurate object detection and semantic segmentation[C]// Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition, Columbus, Jun 20-23, 2014. Washington: IEEE Computer Society, 2014: 580-587.
[2]	EVERINGHAM M, VAN GOOL L, WILLIAMS C K I, et al. The Pascal visual object classes (VOC) challenge[J]. Inter-national Journal of Computer Vision, 2010, 88(2): 303-338.
[3]	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot MultiBox detector[C]// LNCS 9905: Proceedings of the 14th European Conference on Computer Vision, Amsterdam, Oct 8-16, 2016. Cham: Springer, 2016: 21-37.
[4]	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, Jun 27-30, 2016. Washington: IEEE Computer Society, 2016: 779-788.
[5]	LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]// Proceedings of the 2017 IEEE International Conference on Computer Vision, Venice, Oct 22-29, 2017. Washington: IEEE Computer Society, 2017: 2999-3007.
[6]	HE K M, ZHANG X Y, REN S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recogni-tion[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916. DOI URL
[7]	GIRSHICK R. Fast R-CNN[C]// Proceedings of the 2015 IEEE International Conference on Computer Vision, Santiago, Dec 13-16, 2015. Washington: IEEE Computer Society, 2015: 1440-1448.
[8]	REN S, HE K, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[C]// Advances in Neural Information Processing Systems 28, Dec 7-12, 2015. Red Hook: Curran Associates, 2015: 91-99.
[9]	DAI J, LI Y, HE K, et al. R-FCN: object detection via region-based fully convolutional networks[C]// Advances in Neural Information Processing Systems 29, Barcelona, Dec 5-10, 2016. Red Hook: Curran Associates, 2016: 379-387.
[10]	HE K M, GKIOXARI G, DOLLÁR P, et al. Mask R-CNN[C]// Proceedings of the 2017 IEEE International Conference on Computer Vision, Venice, Oct 22-29, 2017. Washington: IEEE Computer Society, 2017: 2980-2988.
[11]	RENSINK R A, ENNS J T. Early completion of occluded objects[J]. Vision Research, 1998, 38(15/16): 2489-2505. DOI URL
[12]	CHEN N, LI M L, YUAN H, et al. Survey of pedestrian detection with occlusion[J]. Complex & Intelligent Systems, 2021, 7(1): 577-587.
[13]	SALEH K, SZÉNÁSI S, VÁMOSSY Z. Occlusion handling in generic object detection: a review[C]// Proceedings of the 2021 IEEE 19th World Symposium on Applied Machine Intelligence and Informatics, Herl'any, Jan 21-23, 2021. Piscataway: IEEE, 2021.
[14]	SHANG M, XIANG D, WANG Z, et al. V2F-Net: explicit decomposition of occluded pedestrian detection[J]. arXiv: 2104.03106, 2021.
[15]	ZHANG S S, BENENSON R, OMRAN M, et al. How far are we from solving pedestrian detection?[C]// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, Jun 27-30, 2016. Washington: IEEE Computer Society, 2016: 1259-1267.
[16]	ZHANG S, WEN L, BIAN X, et al. Occlusion-aware R-CNN: detecting pedestrians in a crowd[C]// LNCS 11207: Proceedings of the 15th European Conference on Computer Vision, Munich, Sep 8-14, 2018. Cham: Springer, 2018: 657-674.
[17]	LIU Y, JING X Y, NIE J H, et al. Context-aware three-dimensional mean-shift with occlusion handling for robust object tracking in RGB-D videos[J]. IEEE Transactions on Multimedia, 2019, 21(3): 664-677. DOI URL
[18]	ONG J, VO B T, VO B N, et al. A Bayesian filter for multi-view 3D multi-object tracking with occlusion handling[J]. arXiv: 2001.04118, 2020.
[19]	ZHAO S C, ZHANG S L, ZHANG L. Towards occlusion handling: object tracking with background estimation[J]. IEEE Transactions on Cybernetics, 2018, 48(7): 2086-2100. DOI URL
[20]	YANG S, LUO P, LOY C C, et al. Faceness-Net: face detec-tion through deep facial part responses[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(8): 1845-1859. DOI URL
[21]	ZENG D, VELDHUIS R, SPREEUWERS L. A survey of face recognition techniques under occlusion[J]. arXiv: 2006.11366, 2020.
[22]	LI A, YUAN Z. SymmNet: a symmetric convolutional neural network for occlusion detection[J]. arXiv: 1807.00959, 2018.
[23]	GILROY S, JONES E, GLAVIN M. Overcoming occlusion in the automotive environment—a review[J]. IEEE Transac-tions on Intelligent Transportation Systems, 2021, 22(1): 23-35.
[24]	REDDY N D, VO M, NARASIMHAN S G. Occlusion-Net: 2D/3D occluded keypoint localization using graph net-works[C]// Proceedings of the 2019 IEEE Conference on Computer Vision and Pattern Recognition, Long Beach, Jun 16-20, 2019. Piscataway: IEEE, 2019: 7326-7335.
[25]	YAN S Y, LIU Q S. Inferring occluded features for fast object detection[J]. Signal Processing, 2015, 110: 188-198. DOI URL
[26]	FAWZI A, FROSSARD P. Measuring the effect of nuisance variables on classifiers[C]// Proceedings of the British Machine Vision Conference 2016, York, Sep 19-22, 2016. Durham: BMVA Press, 2016: 137.
[27]	LIN T Y, MAIRE M, BELONGIE S J, et al. Microsoft COCO: common objects in context[C]// LNCS 8693: Proceedings of the 13th European Conference on Computer Vision, Zurich, Sep 6-12, 2014. Cham: Springer, 2014: 740-755.
[28]	DENG J, DONG W, SOCHER R, et al. ImageNet: a large-scale hierarchical image database[C]// Proceedings of the 2009 IEEE Conference on Computer Vision and Pattern Recognition, Miami, Jun 20-25, 2009. Washington: IEEE Computer Society, 2009: 248-255.
[29]	KUZNETSOVA A, ROM H, ALLDRIN N, et al. The open images dataset v4[J]. International Journal of Computer Vision, 2020, 128(7): 1956-1981. DOI URL
[30]	GEIGER A, LENZ P, URTASUN R. Are we ready for autonomous driving? The KITTI vision benchmark suite[C]// Proceedings of the 2012 IEEE Conference on Computer Vision and Pattern Recognition, Providence, Jun 16-21, 2012. Washington: IEEE Computer Society, 2012: 3354-3361.
[31]	DOLLÁR P, WOJEK C, SCHIELE B, et al. Pedestrian detection: a benchmark[C]// Proceedings of the 2009 IEEE Conference on Computer Vision and Pattern Recognition, Miami, Jun 20-25, 2009. Washington: IEEE Computer Society, 2009: 304-311.
[32]	DOLLÁR P, WOJEK C, SCHIELE B, et al. Pedestrian detec-tion: an evaluation of the state of the art[J]. IEEE Transac-tions on Pattern Analysis and Machine Intelligence, 2012, 34(4): 743-761.
[33]	ZHANG S S, BENENSON R, SCHIELE B. Citypersons: a diverse dataset for pedestrian detection[C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, Jul 21-26, 2017. Washington: IEEE Computer Society, 2017: 4457-4465.
[34]	WANG J, XIE C, ZHANG Z, et al. Detecting semantic parts on partially occluded objects[J]. arXiv:1707.07819, 2017.
[35]	WANG J, ZHANG Z, XIE C, et al. Unsupervised learning of object semantic parts from internal states of CNNs by population encoding[J]. arXiv:1511.06855, 2015.
[36]	SHAO S, ZHAO Z, LI B, et al. Crowdhuman: a benchmark for detecting human in a crowd[J]. arXiv: 1805.00123, 2018.
[37]	FERRYMAN J, SHAHROKNI A. Pets2009: dataset and challenge[C]// Proceedings of the 2009 12th IEEE Interna-tional Workshop on Performance Evaluation of Tracking and Surveillance, Snowbird, Dec 7-9, 2009. Washington: IEEE Computer Society, 2009: 1-6.
[38]	PATINO L, CANE T, VALLEE A, et al. PETS 2016: dataset and challenge[C]// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition Workshops, Las Vegas, Jun 26-Jul 1, 2016. Washington: IEEE Computer Society, 2016: 1-8.
[39]	VIGNESH K, YADAV G K, SETHI A. Abnormal event detection on BMTT-PETS 2017 surveillance challenge[C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops, Honolulu, Jul 21-26, 2017. Washington: IEEE Computer Society, 2017: 2161-2168.
[40]	EHSANI K, MOTTAGHI R, FARHADI A. SeGAN: seg-menting and generating the invisible[C]// Proceedings of the 2018 IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, Jun 18-22, 2018. Washington: IEEE Computer Society, 2018: 6144-6153.
[41]	ADHIKARI B, PELTOMÄKI J, PUURA J, et al. Faster bou-nding box annotation for object detection in indoor scenes[C]// Proceedings of the 7th European Workshop on Visual Information Processing, Tamper, Nov 26-18, 2018. Piscata-way: IEEE, 2018: 1-6.
[42]	SINGH A, SHA J, NARAYAN K S, et al. Bigbird: a large-scale 3D database of object instances[C]// Proceedings of the 2014 IEEE International Conference on Robotics and Automation, Hong Kong, China, May 31-Jun 5, 2014. Pis-cataway: IEEE, 2014: 509-516.
[43]	WANG X L, SHRIVASTAVA A, GUPTA A. A-Fast-RCNN: hard positive generation via adversary for object detection[C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, Jul 21-26, 2017. Washington: IEEE Computer Society, 2017: 2606-2615.
[44]	GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[C]// Advances in Neural Infor-mation Processing Systems 27, Montreal, Dec 8-13, 2014. Red Hook: Curran Associates, 2014: 2672-2680.
[45]	DEVRIES T, TAYLOR G W. Improved regularization of convolutional neural networks with cutout[J]. arXiv:1708. 04552, 2017.
[46]	HINTON G E, SRIVASTAVA N, KRIZHEVSKY A, et al. Improving neural networks by preventing co-adaptation of feature detectors[J]. arXiv:1207.0580, 2012.
[47]	YUN S, HAN D, OH S J, et al. CutMix: regularization strategy to train strong classifiers with localizable features[C]// Proceedings of the 2019 IEEE/CVF International Con-ference on Computer Vision, Seoul, Oct 27-Nov 3, 2019. Piscataway: IEEE, 2019: 6022-6031.
[48]	厍向阳, 李蕊心, 叶鸥. 融合随机擦除和残差注意力网络的行人重识别[J]. 计算机工程与应用, 2022, 58(3): 215-221.
	SHE X Y, LI R X, YE O. Pedestrian re-identification combining random erasing and residual attention network[J]. Computer Engineering and Applications, 2022, 58(3): 215-221.
[49]	TIAN Y L, LUO P, WANG X G, et al. Deep learning strong parts for pedestrian detection[C]// Proceedings of the 2015 IEEE International Conference on Computer Vision, San-tiago, Dec 11-18, 2015. Washington: IEEE Computer Society, 2015: 1904-1912.
[50]	ZHOU C L, YUAN J S. Multi-label learning of part dete-ctors for occluded pedestrian detection[C]// Proceedings of the 2017 IEEE International Conference on Computer Vision, Venice, Oct 22-29, 2017. Washington: IEEE Computer So-ciety, 2017: 3486-3495.
[51]	ZHU Y S, ZHAO C Y, WANG J Q, et al. CoupleNet: coupling global structure with local parts for object detec-tion[C]// Proceedings of the 2017 IEEE International Con-ference on Computer Vision, Venice, Oct 22-29, 2017. Wa-shington: IEEE Computer Society, 2017: 4146-4154.
[52]	LIU T R, LUO W H, MA L, et al. Coupled network for robust pedestrian detection with gated multi-layer feature extraction and deformable occlusion handling[J]. IEEE Trans-actions on Image Processing, 2021, 30: 754-766.
[53]	CHI C, ZHANG S F, XING J L, et al. Relational learning for joint head and human detection[C]// Proceedings of the 34th AAAI Conference on Artificial Intelligence, the 32nd Innovative Applications of Artificial Intelligence Conference, the 10th AAAI Symposium on Educational Advances in Artificial Intelligence, New York, Feb 7-12, 2020. Menlo Park: AAAI, 2020: 10647-10654.
[54]	ZHANG K, XIONG F, SUN P, et al. Double anchor R-CNN for human detection in a crowd[J]. arXiv: 1909.09998, 2019.
[55]	CHU X G, ZHENG A L, ZHANG X Y, et al. Detection in crowded scenes: one proposal, multiple predictions[C]// Pro-ceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, Jun 13-19, 2020. Piscataway: IEEE, 2020: 12211-12220.
[56]	LIN T Y, DOLLÁR P, GIRSHICK R B, et al. Feature pyramid networks for object detection[C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, Jul 21-26, 2017. Washington: IEEE Computer Society, 2017: 2117-2125.
[57]	SHAO X T, WANG Q, YANG W, et al. Multi-scale feature pyramid network: a heavily occluded pedestrian detection network based on ResNet[J]. Sensors, 2021, 21(5): 1820. DOI URL
[58]	YU J H, JIANG Y L, WANG Z Y, et al. UnitBox: an advanced object detection network[C]// Proceedings of the 2016 ACM Conference on Multimedia Conference, Amster-dam, Oct 15-19, 2016. New York: ACM, 2016: 516-520.
[59]	REZATOFIGHI H, TSOI N, GWAK J Y, et al. Generalized intersection over union: a metric and a loss for bounding box regression[C]// Proceedings of the 2019 IEEE Confer-ence on Computer Vision and Pattern Recognition, Long Beach, Jun 16-20, 2019. Piscataway: IEEE, 2019: 658-666.
[60]	ZHENG Z, WANG P, LIU W, et al. Distance-IoU loss: faster and better learning for bounding box regression[C]// Proceedings of the 34th AAAI Conference on Artificial Intelligence, the 32nd Innovative Applications of Artificial Intelligence Conference, the 10th AAAI Symposium on Educational Advances in Artificial Intelligence, New York, Feb 7-12, 2020. Menlo Park: AAAI, 2020: 12993-13000.
[61]	ZHANG Y F, REN W, ZHANG Z, et al. Focal and efficient IOU loss for accurate bounding box regression[J]. arXiv: 2101.08158, 2021.
[62]	WANG X L, XIAO T T, JIANG Y N, et al. Repulsion loss: detecting pedestrians in a crowd[C]// Proceedings of the 2018 IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, Jun 18-22, 2018. Washington: IEEE Computer Society, 2018: 7774-7783.
[63]	阳珊, 王建, 胡莉, 等. 改进RetinaNet的遮挡目标检测算法研究[J]. 计算机工程与应用, 2022, 58(11): 209-214.
	YANG S, WANG J, HU L, et al. Research on occluded object detection by improved RetinaNet[J]. Computer Engineering and Applications, 2022, 58(11): 209-214.
[64]	LUO Z K, FANG Z, ZHENG S X, et al. NMS-Loss: learning with non-maximum suppression for crowded pedestrian detec-tion[C]// Proceedings of the 2021 International Conference on Multimedia Retrieval, Taipei, China, Aug 21-24, 2021. New York: ACM, 2021: 481-485.
[65]	BODLA N, SINGH B, CHELLAPPA R, et al. Soft-NMS-improving object detection with one line of code[C]// Pro-ceedings of the 2017 IEEE International Conference on Computer Vision, Venice, Oct 22-39, 2017. Washington: IEEE Computer Society, 2017: 5561-5569.
[66]	HOSANG J H, BENENSON R, SCHIELE B. Learning non-maximum suppression[C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, Jul 21-26, 2017. Washington: IEEE Computer Society, 2017: 6469-6477.
[67]	LIU S T, HUANG D, WANG Y H. Adaptive NMS: refining pedestrian detection in a crowd[C]// Proceedings of the 2019 IEEE Conference on Computer Vision and Pattern Recogni-tion, Long Beach, Jun 16-20, 2019. Piscataway: IEEE, 2019: 6459-6468.
[68]	HE Y H, ZHU C C, WANG J R, et al. Bounding box regres-sion with uncertainty for accurate object detection[C]// Pro-ceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, Jun 16-20, 2019. Piscataway: IEEE, 2019: 2888-2897.
[69]	HUANG X, GE Z, JIE Z, et al. NMS by representative region: towards crowded pedestrian detection by proposal pairing[C]// Proceedings of the 2020 IEEE Conference on Computer Vision and Pattern Recognition, Seattle, Jun 13-19, 2020. Piscataway: IEEE, 2020: 10750-10759.
[70]	HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]// Proceedings of the 2018 IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, Jun 18-22, 2018. Washington: IEEE Computer Society, 2018: 7132-7141.
[71]	ZHANG S S, YANG J, SCHIELE B. Occluded pedestrian detection through guided attention in CNNs[C]// Proceedings of the 2018 IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, Jun 18-22, 2018. Washington: IEEE Computer Society, 2018: 6995-7003.
[72]	ZHANG S S, CHEN D, YANG J, et al. Guided attention in CNNs for occluded pedestrian detection and re-identification[J]. International Journal of Computer Vision, 2021, 129(6): 1875-1892. DOI URL
[73]	PANG Y W, XIE J, KHAN M H, et al. Mask-guided atten-tion network for occluded pedestrian detection[C]// Procee-dings of the 2019 IEEE/CVF International Conference on Computer Vision, Seoul, Oct 27-Nov 2, 2019. Piscataway: IEEE, 2019: 4966-4974.
[74]	XIE J, PANG Y W, KHAN M H, et al. Mask-guided atten-tion network and occlusion-sensitive hard example mining for occluded pedestrian detection[J]. IEEE Transactions on Image Processing, 2021, 30: 3872-3884. DOI URL
[75]	ZHANG Z S, XIE C H, WANG J Y, et al. DeepVoting: a robust and explainable deep network for semantic part detec-tion under partial occlusion[C]// Proceedings of the 2018 IEEE Conference on Computer Vision and Pattern Recogni-tion, Salt Lake City, Jun 18-22, 2018. Washington: IEEE Computer Society, 2018: 1372-1380.
[76]	SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[J]. arXiv:1409. 1556, 2014.
[77]	KORTYLEWSKI A, LIU Q, WANG H Y, et al. Combining compositional models and deep networks for robust object classification under occlusion[C]// Proceedings of the 2020 IEEE Winter Conference on Applications of Computer Vision, Snowmass Village, Mar 1-5, 2020. Piscataway: IEEE, 2020: 1322-1330.
[78]	XIAO M Q, KORTYLEWSKI A, WU R H, et al. TDAPNet: prototype network with recurrent top-down attention for robust object classification under partial occlusion[J]. arXiv: 1909.03879, 2019.
[79]	KORTYLEWSKI A, HE J, LIU Q, et al. Compositional con-volutional neural networks: a deep architecture with innate robustness to partial occlusion[C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, Jun 13-19, 2020. Piscataway: IEEE, 2020: 8937-8946.
[80]	KORTYLEWSKI A, LIU Q, WANG A T, et al. Compositional convolutional neural networks: a robust and interpretable model for object recognition under occlusion[J]. International Journal of Computer Vision, 2021, 129(3): 736-760. DOI URL
[81]	WANG A T, SUN Y H, KORTYLEWSKI A, et al. Robust object detection under occlusion with context-aware compo-sitionalnets[C]// Proceedings of the 2020 IEEE/CVF Confer-ence on Computer Vision and Pattern Recognition, Seattle, Jun 13-19, 2020. Piscataway: IEEE, 2020: 12642-12651.
[82]	DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An image is worth 16×16 words: Transformers for image reco-gnition at scale[J]. arXiv: 2010.11929, 2020.
[83]	KHAN S, NASEER M, HAYAT M, et al. Transformers in vision: a survey[J]. arXiv: 2101.01169, 2021.

深度学习应用于遮挡目标检测算法综述

Review of Deep Learning Applied to Occluded Object Detection

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 14

参考文献 83

相关文章 15

编辑推荐

Metrics

[1]	安凤平, 李晓薇, 曹翔. 权重初始化-滑动窗口CNN的医学图像分类[J]. 计算机科学与探索, 2022, 16(8): 1885-1897.
[2]	曾凡智, 许露倩, 周燕, 周月霞, 廖俊玮. 面向智慧教育的知识追踪模型研究综述[J]. 计算机科学与探索, 2022, 16(8): 1742-1763.
[3]	刘艺, 李蒙蒙, 郑奇斌, 秦伟, 任小广. 视频目标跟踪算法综述[J]. 计算机科学与探索, 2022, 16(7): 1504-1515.
[4]	赵小明, 杨轶娇, 张石清. 面向深度学习的多模态情感识别研究进展[J]. 计算机科学与探索, 2022, 16(7): 1479-1503.
[5]	夏鸿斌, 肖奕飞, 刘渊. 融合自注意力机制的长文本生成对抗网络模型[J]. 计算机科学与探索, 2022, 16(7): 1603-1610.
[6]	刘雅芬, 郑艺峰, 江铃燚, 李国和, 张文杰. 深度半监督学习中伪标签方法综述[J]. 计算机科学与探索, 2022, 16(6): 1279-1290.
[7]	程卫月, 张雪琴, 林克正, 李骜. 融合全局与局部特征的深度卷积神经网络算法[J]. 计算机科学与探索, 2022, 16(5): 1146-1154.
[8]	钟梦圆, 姜麟. 超分辨率图像重建算法综述[J]. 计算机科学与探索, 2022, 16(5): 972-990.
[9]	裴利沈, 赵雪专. 群体行为识别深度学习方法研究综述[J]. 计算机科学与探索, 2022, 16(4): 775-790.
[10]	许嘉, 韦婷婷, 于戈, 黄欣悦, 吕品. 题目难度评估方法研究综述[J]. 计算机科学与探索, 2022, 16(4): 734-759.
[11]	朱伟杰, 陈莹. 双流时间域信息交互的微表情识别卷积网络[J]. 计算机科学与探索, 2022, 16(4): 950-958.
[12]	张全贵, 胡嘉燕, 王丽. 耦合用户公共特征的单类协同过滤推荐算法[J]. 计算机科学与探索, 2022, 16(3): 637-648.
[13]	姜艺, 胥加洁, 柳絮, 朱俊武. 边缘指导图像修复算法研究[J]. 计算机科学与探索, 2022, 16(3): 669-682.
[14]	邬开俊, 黄涛, 王迪聪, 白晨帅, 陶小苗. 视频异常检测技术研究进展[J]. 计算机科学与探索, 2022, 16(3): 529-540.
[15]	刘利平, 孙建, 高世妍. 单图像盲去模糊方法概述[J]. 计算机科学与探索, 2022, 16(3): 552-564.