目标检测中框回归损失函数的研究

doi:10.3778/j.issn.1002-8331.2009-0450

摘要/Abstract

摘要：

在目标检测中，框回归损失函数的设定直接影响预测框的定位准确性。预测框与目标框的交并比（IOU）被设定为优化预测框的损失函数，但是当两框无重叠面积时无法进行梯度回传。广义的交并比（GIOU）在IOU损失函数的基础上增加非重叠面积部分，将两部分优化项作为损失函数调整预测框位置，解决了无法梯度回传的情况。但当两框是包含关系时，GIOU的第二部分优化项消失，损失函数退化为IOU。为了解决以上问题，提出了一种重新定义的广义交并比损失函数（RGIOU），将非重叠部分面积定义为两框之并减去两框之交，再除以两框形成的最小闭包面积作为第一部分，除以最小闭包面积的平方作为第二部分，利用权重阈值进行加和形成新的损失函数。避免了两框是包含关系时存在的问题，提升了目标检测算法的精度。上述算法在PASCAL VOC 2007以及MS COCO 2014数据集上加以验证。

关键词: 目标检测, 框回归, 交并比（IOU）, 广义的交并比（GIOU）

Abstract:

In object detection, the setting of box regression loss function directly affects the location accuracy of prediction box. The Intersection over Union（IOU） of the prediction frame and the target frame is used to set the loss function of the optimized prediction box, but the gradient return cannot be carried out when the two frames have no overlapping area. Based on the IOU loss function, Generalized Intersection over Union（GIOU） adds a non-overlapping area part to the IOU loss function, and takes the two parts as optimization items to adjust the position of the prediction box, so as to avoid the situation of gradient return. However, when the two boxes are inclusive, the second optimization term of GIOU disappears, GIOU degenerates to IOU. In order to solve the above problems, Redefined Generalized Intersection over Union（RGIOU） loss function is proposed, which redefines the area of non-overlapping part as the sum of two frames minus the intersection of two frames, and then divided by the minimum closure area formed by the two frames as part one, divide by the square of the minimum closure area as the second part, and then a new loss function is formed by adding the weight threshold. It avoids the problem that the two frames are inclusion relations, and improves the accuracy of the object detection algorithm. The algorithm is validated on Pascal VOC 2007 and MS COCO 2014 data sets.

Key words: object detection, bounding box regression, Intersection over Union（IOU）, Generalized Intersection over Union（GIOU）

张翠文，张长伦，何强，王恒友. 目标检测中框回归损失函数的研究[J]. 计算机工程与应用, 2021, 57(20): 97-103.

ZHANG Cuiwen, ZHANG Changlun, HE Qiang, WANG Hengyou. Research on Loss Function of Box Regression in Object Detection[J]. Computer Engineering and Applications, 2021, 57(20): 97-103.

参考文献

[1] 张泽轩，陈虎，吴志红，等.基于深度学习的道路实景行人车辆检测[J].现代计算机（专业版），2018（10）：34-38.
ZHANG Z X，CHEN H，WU Z H，et al.Detection of pedestrian and vehicle based on deep learning[J].Modern Computer，2018（10）：34-38.
[2] 陈超，宣士斌，徐俊格.复杂背景下的行人检测与分割[J].计算机工程与应用，2012，48（30）：177-181.
CHEN C，XUAN S B，XU J G.Pedestrian detection and segmentation under background clutter[J].Computer Engineering and Applications，2012，48（30）：177-181.
[3] 李云鹏，侯凌燕，王超.基于YOLOv3的自动驾驶中运动目标检测[J].计算机工程与设计，2019，40（4）：246-251.
LI Y P，HOU L Y，WANG C.Moving objects detection in automatic driving based on YOLOv3[J].Computer Engineering and Design，2018，40（12）：2812-2819.
[4] 张斌，何明，陈希亮，等.改进DDPG算法在自动驾驶中的应用[J].计算机工程与应用，2019，55（10）：264-270.
ZHANG B，HE M，CHEN X L，et al.Self-driving via improved DDPG algorithm[J].Computer Engineering and Applications，2019，55（10）：264-270.
[5] 戴凤智，魏宝昌，欧阳育星，等.基于深度学习的视频跟踪研究进展综述[J].计算机工程与应用，2019，55（10）：16-29.
DAI F Z，WEI B C，OUYANG Y X，et al.Survey of research progress of video tracking based on deep learning[J].Computer Engineering and Applications，2019，55（10）：16-29.
[6] 袁国武.智能视频监控中的运动目标检测和跟踪算法研究[D].昆明：云南大学，2012.
YUAN G W.Research on moving object detection and tracking algorithm in intelligent video surveillance[D].Kunming：Yunnan University，2012.
[7] 孙彦，丁学文，雷雨婷.基于目标检测模型的人脸识别技术[J].计算机与网络，2019，45（22）：68-71.
SUN Y，DING X W，LEI Y T.Face recognition technology based on object detection model[J].Computer and Network，2019，45（22）：68-71.
[8] 苏煜，山世光，陈熙霖，等.基于全局和局部特征集成的人脸识别[J].软件学报，2010（8）：71-84.
SU Y，SHAN S G，CHEN X L，et al.Integration of global and local feature for face recognition[J].Journal of Software，2010（8）：71-84.
[9] 张丽.车辆视频检测与跟踪系统的算法研究[D].杭州：浙江大学，2003.
ZHANG L.Research on algorithm of vehicle video detection and tracking system[D].Hangzhou：Zhejiang Univesity，2003.
[10] 王夏黎，周明全，耿国华.基于视频检测和颜色的车辆牌照提取方法[J].计算机应用与软件，2005（11）：41-44.
WANG X L，ZHOU M Q，GENG H H.An approach of vehicle plate extract based on visual detection and color information[J].Computer Applications and Software，2005（11）：41-44.
[11] DROLSHAGEN G，BASSANO E，BERNARDI F，et al.Precursor services for a near-earth object segment of ESA’s space situational awareness programme[J].Nature，2012，215：6229.
[12] CHANG W Z，XIAN Z L，JUN D，et al.Infrared ship object segment based on MFFK and MeanShift filtering[J].Laser & Infrared，2010，40（9）：1023-1026.
[13] TSUCHIE S，HIGASHI M.High-quality segmentation of scanned data for industrial object：segment generation from clustered vertices[C]//JSPE Semestrial Meeting，2013.
[14] KOSCHNY D V，DROLSHAGEN G，BOBRINSKY N.The near-earth object segment of ESA’s space situational awareness programme[C]//Making Safety Matter，2010.
[15] KRIZHEVSKY A，SUTSKEVER I，HINTON G.ImageNet classification with deep convolutional neural networks[C]//Proceedings of NIPS，2012.
[16] GIRSHICK R，DONAHUE J，DARRELL T，et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//IEEE Conference on Computer Vision and Pattern Recognition，2014：580-587.
[17] 丁世飞，齐丙娟，谭红艳.支持向量机理论与算法研究综述[J].电子科技大学学报，2011，40（1）：1-10.
DING S F，QI B J，TAN H Y.An review on theory and algorithm support vector machines[J].Journal of University of Electronic Science and Technology of China，2011，40（1）：1-10.
[18] GIRSHICK R.Fast R-CNN[C]//IEEE International Conference on Computer Vision（ICCV），2015.
[19] REN S，HE K，GIRSHICK R，et al.Faster R-CNN：towards real-time object detection with region proposal networks[J].IEEE Transactions on Pattern Analysis & Machine Intelligence，2017，39（6）：1137-1149.
[20] REDMON J，DIVVALA S，GIRSHICK R，et al.You only look once：unified，real-time object detection[C]//IEEE Conference on Computer Vision and Pattern Recognition，2015.
[21] REDMON J，FARHADI A.YOLO9000：better，faster，stronger[C]//IEEE Conference on Computer Vision and Pattern Recognition（CVPR），2016.
[22] REDMON J，FARHADI A.YOLOv3：an incremental improvement[C]//IEEE Conference on Computer Vision and Pattern Recognition，2018.
[23] YU J，JIANG Y，WANG Z，et al.UnitBox：an advanced object detection network[J].arXiv：1608.01471，2016.
[24] REZATOFIGHI H，TSOI N，GWAK J Y，et al.Generalized intersection over union：a metric and a loss for bounding box regression[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition（CVPR），2019.