结合长尾数据解决方法的野生动物目标检测

doi:10.11772/j.issn.1001-9081.2021071279

《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (4): 1284-1291.DOI: 10.11772/j.issn.1001-9081.2021071279

所属专题： CCF第36届中国计算机应用大会 (CCF NCCA 2021)

• CCF第36届中国计算机应用大会 (CCF NCCA 2021) • 上一篇下一篇

结合长尾数据解决方法的野生动物目标检测

蔡前舟¹, 郑伯川²(), 曾祥银², 侯金³

^1.西华师范大学数学与信息学院，四川南充 637009
^2.西华师范大学计算机学院，四川南充 637009
^3.西华师范大学生命科学学院，四川南充 637009

收稿日期:2021-07-16 修回日期:2021-08-23 接受日期:2021-08-27 发布日期:2022-04-28 出版日期:2022-04-10
通讯作者: 郑伯川
作者简介:蔡前舟（1996—），男，四川通江人，硕士研究生，CCF会员，主要研究方向：深度学习、目标检测
曾祥银（1994—），男，四川大竹人，硕士研究生，主要研究方向：机器学习、深度学习
侯金（1995—），男，四川攀枝花人，硕士研究生，主要研究方向：生物多样性保护、可持续发展。
基金资助:
国家自然科学基金资助项目(62176217);四川省科技创新苗子工程项目(2020029);西华师范大学大学生创新创业项目(cxcy2020150)

Wildlife object detection combined with solving method of long-tail data

Qianzhou CAI¹, Bochuan ZHENG²(), Xiangyin ZENG², Jin HOU³

^1.School of Mathematics and Information，China West Normal University，Nanchong Sichuan 637009，China
^2.School of Computer Science，China West Normal University，Nanchong Sichuan 637009，China
^3.College of Life Sciences，China West Normal University，Nanchong Sichuan 637009，China

Received:2021-07-16 Revised:2021-08-23 Accepted:2021-08-27 Online:2022-04-28 Published:2022-04-10
Contact: Bochuan ZHENG
About author:CAI Qianzhou， born in 1996， M. S. candidate. His research interests include deep learning， object detection.
ZENG Xiangyin， born in 1994， M. S. candidate. His research interests include machine learning， deep learning.
HOU Jin， born in 1995， M. S. candidate. His research interests include biodiversity conservation， sustainable development.
Supported by:
National Natural Science Foundation of China(62176217);Program of Sichuan Science and Technology Innovation Seedling Project(2020029);Innovation and Entrepreneurship Project of College Students of China West Normal University(cycx2020150)

摘要/Abstract

摘要：

基于红外相机图像的野生动物目标检测有利于研究和保护野生动物。由于不同种类的野生动物数量差别大，红外相机采集到的野生动物数据集存在种类数量分布不均的长尾数据问题，进而影响目标检测神经网络模型的整体性能提升。针对野生动物的长尾数据导致的目标检测精度低的问题，提出了一种基于两阶段学习和重加权相结合的长尾数据解决方法，并将该方法用于基于YOLOv4-Tiny的野生动物目标检测。首先，采集、标注并构建了一个新的野生动物数据集，该数据集具有明显的长尾数据特征；其次，采用基于迁移学习的两阶段方法训练神经网络，第一阶段在分类损失函数中采用无加权方式进行训练，而在第二阶段提出了两种改进的重加权方法，并以第一阶段所得权重作为预训练权重进行重加权训练；最后，对野生动物测试集进行测试。实验结果表明，在分类损失采用交叉熵损失函数和焦点损失函数下，所提出的长尾数据解决方法达到了60.47%和61.18%的平均精确率均值（mAP），相较于无加权方法在两种损失函数下分别提高了3.30个百分点和5.16个百分点，相较于所提改进的有效样本加权方法在焦点损失函数下提高了2.14个百分点，说明该方法能提升YOLOv4-Tiny网络对具有长尾数据特征的野生动物数据集的目标检测性能。

关键词: 长尾数据, 目标检测, 两阶段学习, 重加权, YOLOv4-Tiny

Abstract:

Wild animal object detection based on infrared camera images is conducive to the research and protection of wild animals. Because of the large difference in the number of different species of wildlife， there is the long-tail data problem of uneven distribution of numbers of species in the wildlife dataset collected by infrared cameras. This problem affects the overall performance improvement of the object detection neural network models. In order to solve the problem of low accuracy of object detection caused by long-tail data of wild animals， a method based on two-stage learning and re-weighting to solve long-tail data was proposed， and the method was applied to wildlife object detection based on YOLOv4-Tiny. Firstly， a new wildlife dataset with obvious long-tail data characteristics was collected， labelled and constructed. Secondly， a two-stage method based on transfer learning was used to train the neural network. In the first stage， the classification loss function was trained without weighting. In the second stage， two improved re-weighting methods were proposed， and the weights obtained in the first stage were used as the pre-training weights for re-weighting training. Finally， the wildlife test set was used to tested. Experimental results showed that the proposed long-tail data solving method achieved 60.47% and 61.18% mAP （mean Average Precision） with cross-entropy loss function and focal loss function as classification loss respectively， which was 3.30 percentage points and 5.16 percentage points higher than that the no-weighting method under the two loss functions， and 2.14 percentage points higher than that of the proposed improved effective sample weighting method under focus loss function. It shows that the proposed method can improve the object detection performance of YOLOv4-Tiny network for wildlife datasets with long-tail data characteristics.

Key words: long-tail data, object detection, two-stage learning, re-weighting, YOLOv4-Tiny

中图分类号:

TP183

蔡前舟, 郑伯川, 曾祥银, 侯金. 结合长尾数据解决方法的野生动物目标检测[J]. 计算机应用, 2022, 42(4): 1284-1291.

Qianzhou CAI, Bochuan ZHENG, Xiangyin ZENG, Jin HOU. Wildlife object detection combined with solving method of long-tail data[J]. Journal of Computer Applications, 2022, 42(4): 1284-1291.

图/表 9

图1 野生动物数据集的种类数量分布

Fig. 1 Distribution of the numbers of species in wildlife dataset

图2 两阶段训练过程

Fig. 2 Process of two-stage training

图3 YOLOv4-Tiny网络结构

Fig. 3 Network structure of YOLOv4-Tiny

图4 交叉熵损失函数的变化图

Fig. 4 Change graph of cross-entropy loss function

图5 焦点损失函数的变化图

Fig. 5 Change graph of focal loss function

表1 本文方法与无加权方法的mAP对比 (%)

Tab. 1 mAP comparison of proposed methods and no-weighting method

阶段	方法	CE loss	Focal loss
阶段	方法	CE loss	$γ = 0.5$	$γ = 1$	$γ = 2$
一阶段	无加权	57.17	56.02	55.43	55.83
	本文加权方法一	58.40	58.33	57.88	57.45
	本文加权方法二	59.12	59.80	59.10	58.01
二阶段	本文加权方法一	59.70	59.59	59.41	58.19
二阶段	本文加权方法二	60.47	61.18	60.68	59.48

表1 本文方法与无加权方法的mAP对比 (%)

Tab. 1 mAP comparison of proposed methods and no-weighting method

阶段	方法	CE loss	Focal loss
阶段	方法	CE loss	$γ = 0.5$	$γ = 1$	$γ = 2$
一阶段	无加权	57.17	56.02	55.43	55.83
	本文加权方法一	58.40	58.33	57.88	57.45
	本文加权方法二	59.12	59.80	59.10	58.01
二阶段	本文加权方法一	59.70	59.59	59.41	58.19
二阶段	本文加权方法二	60.47	61.18	60.68	59.48

图6 不同动物种类的AP对比

Fig. 6 AP comparison of different species

表2 无加权方法与不同加权方法的mAP对比 (%)

Tab. 2 mAP comparison of no-weighting method and different weighting methods

方法	mAP	方法	mAP
YOLOv4-Tiny（无加权）	57.17	有效样本加权	59.04
逆序加权	59.01	LDAMLoss	60.12
逆序平方根加权	58.92	本文加权方法二	61.18

图7 不同加权方法的部分野生动物检测结果

Fig. 7 Some wildlife detection results of different weighting methods

参考文献 33

1	NGUYEN H， MACLAGAN S J， NGUYEN T D， et al. Animal recognition and identification with deep convolutional neural networks for automated wildlife monitoring［C］// Proceedings of the 2017 IEEE International Conference on Data Science and Advanced Analytics. Piscataway： IEEE， 2017：40-49. 10.1109/dsaa.2017.31
2	CHEN G B， HAN T X， HE Z H， et al. Deep convolutional neural network based species recognition for wild animal monitoring［C］// Proceedings of the 2014 IEEE International Conference on Image Processing. Piscataway： IEEE， 2014：858-862. 10.1109/icip.2014.7025172
3	GOMEZ VILLA A， SALAZAR A， VARGAS F. Towards automatic wild animal monitoring： identification of animal species in camera-trap images using very deep convolutional neural networks［J］. Ecological Informatics， 2017， 41：24-32. 10.1016/j.ecoinf.2017.07.004
4	ZHU C B， LI T H， LI G. Towards automatic wild animal detection in low quality camera-trap images using two-channeled perceiving residual pyramid networks［C］// Proceedings of the 2017 IEEE International Conference on Computer Vision Workshops. Piscataway： IEEE， 2017：2860-2864. 10.1109/iccvw.2017.337
5	OUYANG W L， WANG X G， ZHANG C， et al. Factors in finetuning deep model for object detection with long-tail distribution［C］// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2016：864-873. 10.1109/cvpr.2016.100
6	LIN T Y， MAIRE M， BELONGIE S， et al. Microsoft COCO： common objects in context［C］// Proceeding of the 2014 European Conference on Computer Vision， LNCS 8693. Cham： Springer， 2014：740-755.
7	DENG J， DONG W， SOCHER R， et al. ImageNet： a large-scale hierarchical image database［C］// Proceeding of the 2009 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2009： 248-255. 10.1109/cvpr.2009.5206848
8	TORRALBA A， FERGUS R， FREEMAN W T. 80 million tiny images： a large data set for nonparametric object and scene recognition［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2008， 30（11）：1958-1970. 10.1109/tpami.2008.128
9	POUYANFAR S， TAO Y D， MOHAN A， et al. Dynamic sampling in convolutional neural networks for imbalanced data classification［C］// Proceeding of the 2018 IEEE Conference on Multimedia Information Processing and Retrieval. Piscataway： IEEE， 2018： 112-117. 10.1109/mipr.2018.00027
10	HE H B， GARCIA E A. Learning from imbalanced data［J］. IEEE Transactions on Knowledge and Data Engineering， 2009， 21（9）：1263-1284. 10.1109/tkde.2008.239
11	王俊红，闫家荣. 基于欠采样和代价敏感的不平衡数据分类算法［J］. 计算机应用， 2021， 41（1）：48-52.
	WANG J H， YAN J R. Classification algorithm based on undersampling and cost-sensitiveness for unbalanced data［J］. Journal of Computer Applications， 2021， 41（1）：48-52.
12	HUANG C， LI Y N， LOY C C， et al. Learning deep representation for imbalanced classification［C］// Proceeding of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2016：5375-5384. 10.1109/cvpr.2016.580
13	CUI Y， JIA M L， LIN T Y， et al. Class-balanced loss based on effective number of samples［C］// Proceeding of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2019：9260-9269. 10.1109/cvpr.2019.00949
14	CAO K D， WEI C， GAIDON A， et al. Learning imbalanced datasets with label-distribution-aware margin loss［C］// Proceedings of the 33rd International Conference on Neural Information Processing Systems. Red Hook， NY： Curran Associates Inc.， 2019： 1567-1578.
15	姚佳奇，徐正国，燕继坤，等. WPLoss：面向类别不平衡数据的加权成对损失［J］. 计算机应用研究， 2021， 38（3）：702-704， 709.
	YAO J Q， XU Z G， YAN J K， et al. WPLoss： weighted pairwise loss for class-imbalanced datasets［J］. Application Research of Computers， 2021， 38（3）：702-704， 709.
16	XIANG L Y， DING G G， HAN J G. Learning from multiple experts： self-paced knowledge distillation for long-tailed classification［C］// Proceeding of the 2020 European Conference on Computer Vision， LNCS 12350. Cham： Springer， 2020：247-263.
17	ZHANG S Y， CHEN C， HU X Y， et al. Balanced knowledge distillation for long-tailed learning［EB/OL］. （2021-04-21）［2021-05-13］..
18	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. Piscataway： IEEE， 2016：779-788. 10.1109/cvpr.2016.91
19	REDMON J， FARHADI A. YOLO9000： better， faster， stronger［C］// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2017：6517-6525. 10.1109/cvpr.2017.690
20	REDMON R， FARHADI A. YOLOv3： an incremental improvement［EB/OL］. （2018-04-08）［2021-05-13］..
21	JIANG Z C， ZHAO L Q， LI S Y， et al. Real-time object detection method based on improved YOLOv4-tiny［EB/OL］. （2020-12-02）［2021-05-13］..
22	WANG Y X， RAMANAN D， HEBERT M. Learning to model the tail［C］// Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook， NY： Curran Associates Inc.， 2017： 7032-7042. 10.1109/cvpr.2017.323
23	CHAWLA N V， BOWYER K W， HALL L O， et al. SMOTE： synthetic minority over-sampling technique［J］. Journal of Artificial Intelligence Research， 2002， 16：321-357. 10.1613/jair.953
24	HAN H， WANG W Y， MAO B H. Borderline-SMOTE： a new over-sampling method in imbalanced data sets learning［C］// Proceedings of the 2005 International Conference on Intelligent Computing， LNCS 3644. Berlin： Springer， 2005：878-887.
25	BUNKHUMPORNPAT C， SINAPIROMSARAN K， LURSINSAP C. Safe-Level-SMOTE： safe-level-synthetic minority over-sampling technique for handling the class imbalanced problem［C］// Proceedings of the 2009 Pacific-Asia Conference on Knowledge Discovery and Data Mining. Berlin： Springer， 2009： 475-482. 10.1007/978-3-642-01307-2_43
26	LIU X Y， WU J X， ZHOU Z H. Exploratory undersampling for class-imbalance learning［J］. IEEE Transactions on Systems， Man， and Cybernetics， Part B （Cybernetics）， 2009， 39（2）：539-550. 10.1109/tsmcb.2008.2007853
27	MIKOLOV T， SUTSKEVER I， CHEN K， et al. Distributed representations of words and phrases and their compositionality［C］// Proceedings of the 26th International Conference on Neural Information Processing Systems. Red Hook， NY： Curran Associates Inc.， 2013：3111-3119.
28	LIN T Y， GOYAL P， GIRSHICK R， et al. Focal loss for dense object detection［C］// Proceeding of the 2017 IEEE International Conference on Computer Vision. Piscataway： IEEE， 2017：2999-3007. 10.1109/iccv.2017.324
29	LIU Z W， MIAO Z Q， ZHAN X H， et al. Large-scale long-tailed recognition in an open world［C］// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2019： 2532-2541. 10.1109/cvpr.2019.00264
30	KANG B Y， XIE S N， ROHRBACH M， et al. Decoupling representation and classifier for long-tailed recognition［EB/OL］. （2020-02-19）［2021-05-13］..
31	ZHOU B Y， CUI Q， WEI X S， et al. BBN： bilateral-branch network with cumulative learning for long-tailed visual recognition［C］// Proceeding of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2020：9716-9725. 10.1109/cvpr42600.2020.00974
32	ZHENG Z H， WANG P， LIU W， et al. Distance-IoU loss： faster and better learning for bounding box regression［C］// Proceeding of the 34th AAAI Conference on Artificial intelligence. Palo Alto， CA： AAAI Press， 2020：12993-13000. 10.1609/aaai.v34i07.6999
33	REZATOFIGHI H， TSOI N， GWAK J， et al. Generalized intersection over union： a metric and a loss for bounding box regression［C］// Proceeding of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2019：658-666. 10.1109/cvpr.2019.00075

[1]	潘烨新, 杨哲. 基于多级特征双向融合的小目标检测优化模型[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2871-2877.
[2]	李烨恒, 罗光圣, 苏前敏. 基于改进YOLOv5的Logo检测算法[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2580-2587.
[3]	张英俊, 李牛牛, 谢斌红, 张睿, 陆望东. 课程学习指导下的半监督目标检测框架[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2326-2333.
[4]	姬张建, 杜娜. 基于改进VariFocalNet的微小目标检测[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2200-2207.
[5]	徐松, 张文博, 王一帆. 基于时空信息的轻量视频显著性目标检测网络[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2192-2199.
[6]	孙逊, 冯睿锋, 陈彦如. 基于深度与实例分割融合的单目3D目标检测方法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2208-2215.
[7]	刘越, 刘芳, 武奥运, 柴秋月, 王天笑. 基于自注意力机制与图卷积的3D目标检测网络[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1972-1977.
[8]	邓亚平, 李迎江. YOLO算法及其在自动驾驶场景中目标检测综述[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1949-1958.
[9]	耿焕同, 刘振宇, 蒋骏, 范子辰, 李嘉兴. 基于改进YOLOv8的嵌入式道路裂缝检测算法[J]. 《计算机应用》唯一官方网站, 2024, 44(5): 1613-1618.
[10]	邹博士, 杨铭, 宗辰辰, 谢明昆, 黄圣君. 基于负学习的样本重加权鲁棒学习方法[J]. 《计算机应用》唯一官方网站, 2024, 44(5): 1479-1484.
[11]	李鸿天, 史鑫昊, 潘卫国, 徐成, 徐冰心, 袁家政. 融合多尺度和注意力机制的小样本目标检测[J]. 《计算机应用》唯一官方网站, 2024, 44(5): 1437-1444.
[12]	宋霄罡, 张冬冬, 张鹏飞, 梁莉, 黑新宏. 面向复杂施工环境的实时目标检测算法[J]. 《计算机应用》唯一官方网站, 2024, 44(5): 1605-1612.
[13]	李宗禹, 强思维, 郭晓波, 朱振峰. 重加权的对抗变分自编码器及其在工业因果效应估计中的应用[J]. 《计算机应用》唯一官方网站, 2024, 44(4): 1099-1106.
[14]	陈天华, 朱家煊, 印杰. 基于注意力机制的鸟类识别算法[J]. 《计算机应用》唯一官方网站, 2024, 44(4): 1114-1120.
[15]	王伟, 赵春辉, 唐心瑶, 席刘钢. 自适应地平线约束下的车辆三维检测[J]. 《计算机应用》唯一官方网站, 2024, 44(3): 909-915.

结合长尾数据解决方法的野生动物目标检测

Wildlife object detection combined with solving method of long-tail data

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 33

相关文章 15

编辑推荐

Metrics