《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (8): 2401-2406.DOI: 10.11772/j.issn.1001-9081.2021060950
所属专题: 人工智能
收稿日期:
2021-06-07
修回日期:
2021-08-17
接受日期:
2021-09-02
发布日期:
2022-08-09
出版日期:
2022-08-10
通讯作者:
余贞侠
作者简介:
程南江(1997—),男,四川内江人,硕士研究生,主要研究方向:机器学习、计算机视觉;基金资助:
Nanjiang CHENG1,2, Zhenxia YU1(), Lin CHEN2, Hezhe QIAO2
Received:
2021-06-07
Revised:
2021-08-17
Accepted:
2021-09-02
Online:
2022-08-09
Published:
2022-08-10
Contact:
Zhenxia YU
About author:
CHENG Nanjiang, born in 1997, M. S. candidate. His research interests include machine learning, computer vision.Supported by:
摘要:
当前行人属性识别(PAR)公开数据集中属性标注繁杂且采集场景多样,各数据集中行人属性差异较大,进而导致公开数据库已有的标记信息数据难以直接应用到PAR实际问题中。针对上述问题,提出一种基于领域自适应的多源多标签PAR方法。首先通过领域自适应方法对样本进行特征对齐完成多个数据集之间的统一风格转换;接着提出多属性one-hot编码加权算法,将多数据集中共有属性的标签对齐;最后结合多标签半监督损失函数,进行跨数据集联合训练以提高属性识别准确率。通过所提出的特征对齐和标签对齐算法,可有效解决PAR多数据集中属性异构性问题。将三个行人属性数据集PETA、RAPv1和RAPv2分别与PA-100K数据集对齐后的实验结果表明,所提出的方法对比StrongBaseline在平均准确率上分别提升了1.22、1.62和1.53个百分点,说明该方法在跨数据集PAR中具有一定的优势。
中图分类号:
程南江, 余贞侠, 陈琳, 乔贺辙. 基于领域自适应的多源多标签行人属性识别[J]. 计算机应用, 2022, 42(8): 2401-2406.
Nanjiang CHENG, Zhenxia YU, Lin CHEN, Hezhe QIAO. Multi-source and multi-label pedestrian attribute recognition based on domain adaptation[J]. Journal of Computer Applications, 2022, 42(8): 2401-2406.
方法 | mA | Acc | Prec | Rec | F1 |
---|---|---|---|---|---|
HPNet[ | 74.21 | 72.19 | 82.97 | 82.09 | 82.53 |
PGDM[ | 74.95 | 73.08 | 84.36 | 82.24 | 83.29 |
CoCNN[ | 80.56 | 78.30 | 89.49 | 84.36 | 86.85 |
ALM[ | 80.68 | 77.08 | 84.21 | 88.84 | 86.46 |
MT-CAS[ | 77.20 | 78.09 | 88.46 | 84.86 | 86.62 |
AR-BiFPN[ | 81.45 | 79.13 | 86.24 | 89.46 | 87.94 |
StrongBaseline[ | 80.50 | 78.84 | 87.24 | 87.12 | 86.78 |
PETA→PA-100K | 81.72 | 78.98 | 87.01 | 87.46 | 87.23 |
RAPv1→PA-100K | 82.12 | 79.38 | 86.67 | 88.04 | 87.35 |
RAPv2→PA-100K | 82.03 | 79.59 | 87.52 | 88.38 | 87.95 |
表1 在PA-100K数据集上的不同方法结果对比 ( %)
Tab. 1 Result comparison of different methods on PA-100K dataset
方法 | mA | Acc | Prec | Rec | F1 |
---|---|---|---|---|---|
HPNet[ | 74.21 | 72.19 | 82.97 | 82.09 | 82.53 |
PGDM[ | 74.95 | 73.08 | 84.36 | 82.24 | 83.29 |
CoCNN[ | 80.56 | 78.30 | 89.49 | 84.36 | 86.85 |
ALM[ | 80.68 | 77.08 | 84.21 | 88.84 | 86.46 |
MT-CAS[ | 77.20 | 78.09 | 88.46 | 84.86 | 86.62 |
AR-BiFPN[ | 81.45 | 79.13 | 86.24 | 89.46 | 87.94 |
StrongBaseline[ | 80.50 | 78.84 | 87.24 | 87.12 | 86.78 |
PETA→PA-100K | 81.72 | 78.98 | 87.01 | 87.46 | 87.23 |
RAPv1→PA-100K | 82.12 | 79.38 | 86.67 | 88.04 | 87.35 |
RAPv2→PA-100K | 82.03 | 79.59 | 87.52 | 88.38 | 87.95 |
方法 | RAPv1→PA-100K | RAPv2→PA-100K | PETA→PA-100K | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
mA | Acc | Prec | Rec | F1 | mA | Acc | Prec | Rec | F1 | mA | Acc | Prec | Rec | F1 | |
ResNet-50 | 80.50 | 78.84 | 87.12 | 86.78 | 80.50 | 87.24 | 86.78 | 80.50 | 78.84 | 87.12 | 86.78 | ||||
ResNet-50+LAM | 87.30 | 78.36 | 87.68 | 87.05 | 79.05 | 87.30 | |||||||||
ResNet-50+LAM+FAM | 82.12 | 79.38 | 86.67 | 88.04 | 87.35 | 82.03 | 79.59 | 88.38 | 87.95 | 81.72 | 87.01 | 87.46 | 87.23 |
表2 消融实验结果 ( %)
Tab. 2 Results of Ablation experiments
方法 | RAPv1→PA-100K | RAPv2→PA-100K | PETA→PA-100K | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
mA | Acc | Prec | Rec | F1 | mA | Acc | Prec | Rec | F1 | mA | Acc | Prec | Rec | F1 | |
ResNet-50 | 80.50 | 78.84 | 87.12 | 86.78 | 80.50 | 87.24 | 86.78 | 80.50 | 78.84 | 87.12 | 86.78 | ||||
ResNet-50+LAM | 87.30 | 78.36 | 87.68 | 87.05 | 79.05 | 87.30 | |||||||||
ResNet-50+LAM+FAM | 82.12 | 79.38 | 86.67 | 88.04 | 87.35 | 82.03 | 79.59 | 88.38 | 87.95 | 81.72 | 87.01 | 87.46 | 87.23 |
属性 | 识别准确率/% | 提升百分点 | |
---|---|---|---|
StrongBaseline | RAPv1→PA-100K | ||
Hat | 66.16 | 76.84 | 10.68 |
Glasses | 81.23 | 87.07 | 5.84 |
ShortSleeve | 90.71 | 91.61 | 0.90 |
Trousers | 89.85 | 91.92 | 2.07 |
Skirt&Dress | 82.78 | 86.48 | 3.70 |
boots | 53.47 | 54.86 | 1.39 |
HandBag | 75.93 | 81.03 | 5.10 |
ShoulderBag | 76.43 | 76.07 | -0.36 |
Backpack | 79.75 | 82.62 | 2.87 |
Female | 90.61 | 92.37 | 1.76 |
表3 在RAPv1→PA-100K上的共有属性识别准确率对比
Tab. 3 Comparison of common attribute recognition accuracy on RAPv1→PA-100K
属性 | 识别准确率/% | 提升百分点 | |
---|---|---|---|
StrongBaseline | RAPv1→PA-100K | ||
Hat | 66.16 | 76.84 | 10.68 |
Glasses | 81.23 | 87.07 | 5.84 |
ShortSleeve | 90.71 | 91.61 | 0.90 |
Trousers | 89.85 | 91.92 | 2.07 |
Skirt&Dress | 82.78 | 86.48 | 3.70 |
boots | 53.47 | 54.86 | 1.39 |
HandBag | 75.93 | 81.03 | 5.10 |
ShoulderBag | 76.43 | 76.07 | -0.36 |
Backpack | 79.75 | 82.62 | 2.87 |
Female | 90.61 | 92.37 | 1.76 |
属性 | 识别准确率/% | 提升百分点 | |
---|---|---|---|
StrongBaseline | RAPv2→PA-100K | ||
Hat | 66.16 | 74.63 | 8.47 |
Glasses | 81.23 | 87.48 | 6.25 |
ShortSleeve | 90.71 | 91.35 | 0.64 |
Trousers | 89.85 | 91.83 | 1.98 |
Skirt&Dress | 82.78 | 87.59 | 4.81 |
boots | 53.47 | 56.32 | 2.85 |
HandBag | 75.93 | 81.03 | 5.10 |
ShoulderBag | 76.43 | 76.08 | -0.35 |
Backpack | 79.75 | 78.59 | -1.16 |
Female | 90.61 | 92.48 | 1.87 |
表4 在RAPv2→PA-100K上的共有属性识别准确率对比
Tab. 4 Comparison of common attribute recognition accuracy on RAPv2→PA-100K
属性 | 识别准确率/% | 提升百分点 | |
---|---|---|---|
StrongBaseline | RAPv2→PA-100K | ||
Hat | 66.16 | 74.63 | 8.47 |
Glasses | 81.23 | 87.48 | 6.25 |
ShortSleeve | 90.71 | 91.35 | 0.64 |
Trousers | 89.85 | 91.83 | 1.98 |
Skirt&Dress | 82.78 | 87.59 | 4.81 |
boots | 53.47 | 56.32 | 2.85 |
HandBag | 75.93 | 81.03 | 5.10 |
ShoulderBag | 76.43 | 76.08 | -0.35 |
Backpack | 79.75 | 78.59 | -1.16 |
Female | 90.61 | 92.48 | 1.87 |
属性 | 识别准确率/% | 提升百分点 | |
---|---|---|---|
StrongBaseline | PETA→PA-100K | ||
Hat | 66.16 | 73.47 | 7.31 |
Glasses | 81.23 | 85.39 | 4.16 |
ShortSleeve | 90.71 | 91.38 | 0.67 |
UpperLogo | 78.31 | 82.77 | 4.46 |
UpperPlaid | 77.3 | 80.17 | 2.87 |
Trousers | 89.85 | 91.31 | 1.46 |
Shorts | 90.83 | 91.42 | 0.59 |
Skirt&Dress | 82.78 | 85.43 | 2.65 |
Backpack | 79.75 | 80.14 | 0.39 |
Female | 90.61 | 91.22 | 0.61 |
AgeOver60 | 76.89 | 80.12 | 3.23 |
表5 在PETA→PA-100K上的共有属性识别准确率对比
Tab. 5 Comparison of common attribute recognition accuracy on PETA→PA-100K
属性 | 识别准确率/% | 提升百分点 | |
---|---|---|---|
StrongBaseline | PETA→PA-100K | ||
Hat | 66.16 | 73.47 | 7.31 |
Glasses | 81.23 | 85.39 | 4.16 |
ShortSleeve | 90.71 | 91.38 | 0.67 |
UpperLogo | 78.31 | 82.77 | 4.46 |
UpperPlaid | 77.3 | 80.17 | 2.87 |
Trousers | 89.85 | 91.31 | 1.46 |
Shorts | 90.83 | 91.42 | 0.59 |
Skirt&Dress | 82.78 | 85.43 | 2.65 |
Backpack | 79.75 | 80.14 | 0.39 |
Female | 90.61 | 91.22 | 0.61 |
AgeOver60 | 76.89 | 80.12 | 3.23 |
1 | LAYNE R, HOSPEDALES T, GONG S G. Person re-identification by attributes [C]// Proceedings of the 2012 British Machine Vision Conference. Durham: BMVA Press, 2012: No.24. 10.5244/c.26.24 |
2 | WANG J Y, ZHU X T, GONG S G, et al. Transferable joint attribute-identity deep learning for unsupervised person re-identification [C]// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 2275-2284. 10.1109/cvpr.2018.00242 |
3 | KUMAR N, BERG A C, BELHUMEUR P N, et al. Attribute and simile classifiers for face verification [C]// Proceedings of the 2009 IEEE 12th International Conference on Computer Vision. Piscataway: IEEE, 2009: 365-372. 10.1109/iccv.2009.5459250 |
4 | FERIS R, BOBBITT R, BROWN L, et al. Attribute-based people search: lessons learnt from a practical surveillance system [C]// Proceedings of the 4th ACM International Conference on Multimedia Retrieval. New York: ACM, 2014: 153-160. 10.1145/2578726.2578732 |
5 | LIN Y T, ZHENG L, ZHENG Z D, et al. Improving person re-identification by attribute and identity learning[J]. Pattern Recognition, 2019, 95: 151-161. 10.1016/j.patcog.2019.06.006 |
6 | 马路宽.行人属性识别研究[J].现代计算机, 2021(4): 89-93. 10.3969/j.issn.1007-1423.2021.04.018 |
MA L K. Research on pedestrian attribute recognition[J]. Modern Computer, 2021(4): 89-93. 10.3969/j.issn.1007-1423.2021.04.018 | |
7 | LI D W, CHEN X T, HUANG K Q. Multi-attribute learning for pedestrian attribute recognition in surveillance scenarios [C]// Proceedings of the 3rd IAPR Asian Conference on Pattern Recognition. Piscataway: IEEE, 2015: 111-115. 10.1109/acpr.2015.7486476 |
8 | SUDOWE P, SPITZER H, LEIBE B. Person attribute recognition with a jointly-trained holistic CNN model [C]// Proceedings of the 2015 IEEE International Conference on Computer Vision Workshops. Piscataway: IEEE, 2015: 329-337. 10.1109/iccvw.2015.51 |
9 | ABDULNABI A H, WANG G, LU J W, et al. Multi-task CNN model for attribute prediction[J]. IEEE Transactions on Multimedia, 2015, 17(11): 1949-1959. 10.1109/tmm.2015.2477680 |
10 | LI D W, CHEN X T, ZHANG Z, et al. Pose guided deep model for pedestrian attribute recognition in surveillance scenarios [C]// Proceedings of the 2018 IEEE International Conference on Multimedia and Expo. Piscataway: IEEE, 2018: 1-6. 10.1109/icme.2018.8486604 |
11 | SARAFIANOS N, XU X, KAKADIARIS I A. Deep imbalanced attribute classification using visual attention aggregation [C]// Proceedings of the 2018 European Conference on Computer Vision, LNCS 11215. Cham: Springer, 2018: 708-725. |
12 | BOURDEV L, MAJI S, MALIK J. Describing people: a poselet-based approach to attribute classification [C]// Proceedings of the 2011 International Conference on Computer Vision. Piscataway: IEEE, 2011: 1543-1550. 10.1109/iccv.2011.6126413 |
13 | LIU X H, ZHAO H Y, TIAN M Q, et al. HydraPlus-Net: attentive deep features for pedestrian analysis [C]// Proceedings of the 2017 IEEE International Conference on Computer Vision. Piscataway: IEEE, 2017: 350-359. 10.1109/iccv.2017.46 |
14 | WU M D, HUANG D, GUO Y F, et al. Distraction-aware feature learning for human attribute recognition via coarse-to-fine attention mechanism [C]// Proceedings of the 34th AAAI Conference on Artificial Intelligence. Palo Alto, CA: AAAI Press, 2020: 12394-12401. 10.1609/aaai.v34i07.6925 |
15 | YAGHOUBI E, BORZA D, NEVES J, et al. An attention-based deep learning model for multiple pedestrian attributes recognition[J]. Image and Vision Computing, 2020, 102: No.103981. 10.1016/j.imavis.2020.103981 |
16 | 吴锐,刘宇,冯凯.基于双域自注意力机制的行人属性识别[J].计算机应用, 2021, 41(2): 372-378. |
WU R, LIU Y, FENG K. Pedestrian attribute recognition based on two-domain self-attention mechanism[J]. Journal of Computer Applications, 2021, 41(2): 372-378. | |
17 | HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition [C]// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2016, 770-778. 10.1109/cvpr.2016.90 |
18 | DENG Y B, LUO P, LOY C C, et al. Pedestrian attribute recognition at far distance [C]// Proceedings of the 22nd ACM International Conference on Multimedia. New York: ACM, 2014: 789-792. 10.1145/2647868.2654966 |
19 | LI D W, ZHANG Z, CHEN X T, et al. A richly annotated dataset for pedestrian attribute recognition[EB/OL]. (2016-04-27) [2021-08-15]. . 10.1109/tip.2018.2878349 |
20 | LI D W, ZHANG Z, CHEN X T, et al. A richly annotated pedestrian dataset for person retrieval in real surveillance scenarios[J]. IEEE Transactions on Image Processing, 2019, 28(4): 1575-1590. 10.1109/tip.2018.2878349 |
21 | 刘士豪,胡学敏,姜博厚,等.基于生成对抗双网络的虚拟到真实驾驶场景的视频翻译模型[J].计算机应用, 2020, 40(6): 1621-1626. |
LIU S H, HU X M, JIANG B H, et al. Video translation model from virtual to real driving scenes based on generative adversarial dual networks[J]. Journal of Computer Applications, 2020, 40(6): 1621-1626. | |
22 | ZHU J Y, PARK T, ISOLA P, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks [C]// Proceedings of the 2017 IEEE International Conference on Computer Vision. Piscataway: IEEE, 2017: 2242-2251. 10.1109/iccv.2017.244 |
23 | HAN K, WANG Y H, SHU H, et al. Attribute aware pooling for pedestrian attribute recognition [C]// Proceedings of the 28th Joint Conference on Artificial Intelligence. California: ijcai.org, 2019: 2456-2462. 10.24963/ijcai.2019/341 |
24 | TANG C F, SHENG L, ZHANG Z X, et al. Improving pedestrian attribute recognition with weakly-supervised multi-scale attribute-specific localization [C]// Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2019: 4996-5005. 10.1109/iccv.2019.00510 |
25 | ZENG H T, AI H Z, ZHUANG Z J, et al. Multi-task learning via co-attentive sharing for pedestrian attribute recognition [C]// Proceedings of the 2020 IEEE International Conference on Multimedia and Expo. Piscataway: IEEE, 2020: 1-6. 10.1109/icme46284.2020.9102757 |
26 | MOGHADDAM M, CHARMI M, HASSANPOOR H. Jointly human semantic parsing and attribute recognition with feature pyramid structure in EfficientNets[J]. IET Image Processing, 2021, 15(10): 2281-2291. |
27 | JIA J, HUANG H J, YANG W J, et al. Rethinking of pedestrian attribute recognition: realistic datasets with efficient method [EB/OL]. (2020-05-26) [2021-08-15]. . |
[1] | 李顺勇, 李师毅, 胥瑞, 赵兴旺. 基于自注意力融合的不完整多视图聚类算法[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2696-2703. |
[2] | 黄云川, 江永全, 黄骏涛, 杨燕. 基于元图同构网络的分子毒性预测[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2964-2969. |
[3] | 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. |
[4] | 李云, 王富铕, 井佩光, 王粟, 肖澳. 基于不确定度感知的帧关联短视频事件检测方法[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2903-2910. |
[5] | 秦璟, 秦志光, 李发礼, 彭悦恒. 基于概率稀疏自注意力神经网络的重性抑郁疾患诊断[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2970-2974. |
[6] | 王熙源, 张战成, 徐少康, 张宝成, 罗晓清, 胡伏原. 面向手术导航3D/2D配准的无监督跨域迁移网络[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2911-2918. |
[7] | 张春雪, 仇丽青, 孙承爱, 荆彩霞. 基于两阶段动态兴趣识别的购买行为预测模型[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2365-2371. |
[8] | 刘禹含, 吉根林, 张红苹. 基于骨架图与混合注意力的视频行人异常检测方法[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2551-2557. |
[9] | 顾焰杰, 张英俊, 刘晓倩, 周围, 孙威. 基于时空多图融合的交通流量预测[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2618-2625. |
[10] | 石乾宏, 杨燕, 江永全, 欧阳小草, 范武波, 陈强, 姜涛, 李媛. 面向空气质量预测的多粒度突变拟合网络[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2643-2650. |
[11] | 赵宇博, 张丽萍, 闫盛, 侯敏, 高茂. 基于改进分段卷积神经网络和知识蒸馏的学科知识实体间关系抽取[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2421-2429. |
[12] | 陈虹, 齐兵, 金海波, 武聪, 张立昂. 融合1D-CNN与BiGRU的类不平衡流量异常检测[J]. 《计算机应用》唯一官方网站, 2024, 44(8): 2493-2499. |
[13] | 吴筝, 程志友, 汪真天, 汪传建, 王胜, 许辉. 基于深度学习的患者麻醉复苏过程中的头部运动幅度分类方法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2258-2263. |
[14] | 王东炜, 刘柏辰, 韩志, 王艳美, 唐延东. 基于低秩分解和向量量化的深度网络压缩方法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 1987-1994. |
[15] | 李欢欢, 黄添强, 丁雪梅, 罗海峰, 黄丽清. 基于多尺度时空图卷积网络的交通出行需求预测[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2065-2072. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||