《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (4): 1285-1292.DOI: 10.11772/j.issn.1001-9081.2024050566
霍丽娜1, 薛乐仁2, 戴钰俊2, 赵新宇2, 王世行2, 王威1()
收稿日期:
2024-05-09
修回日期:
2024-08-29
接受日期:
2024-09-13
发布日期:
2024-09-27
出版日期:
2025-04-10
通讯作者:
王威
作者简介:
霍丽娜(1982—),女,河北平山人,副教授,博士,主要研究方向:计算机视觉、多模态学习、深度学习基金资助:
Lina HUO1, Leren XUE2, Yujun DAI2, Xinyu ZHAO2, Shihang WANG2, Wei WANG1()
Received:
2024-05-09
Revised:
2024-08-29
Accepted:
2024-09-13
Online:
2024-09-27
Published:
2025-04-10
Contact:
Wei WANG
About author:
HUO Lina, born in 1982, Ph. D., associate professor. Her research interests include computer vision, multi-modal learning, deep learning.Supported by:
摘要:
步态识别旨在通过人们的步行姿态进行身份识别。针对步态识别中有效感受野(ERF)与人体轮廓区域匹配不佳的问题,提出一种基于空洞卷积的步态识别方法DilatedGait。首先,采用空洞卷积扩大神经元感受野,缓解下采样和模型深度增加导致的分辨率下降,以提高轮廓结构的辨识度;其次,提出扩张重参数化模块(DRM),通过重参数化方法融合多尺度卷积核参数,优化ERF聚焦范围,使模型捕获更多的全局上下文信息;最后,通过特征映射提取判别性步态特征。在户外数据集Gait3D和GREW上的实验结果表明,对比目前的先进方法GaitBase,DilatedGait在Gait3D的Rank-1和平均逆负惩罚(mINP)上分别提升了9.0和14.2个百分点,在GREW的Rank-1和Rank-5上分别提升了11.6和8.8个百分点。可见,DilatedGait消除了复杂协变量带来的不利影响,能进一步提升户外场景下步态识别的准确率。
中图分类号:
霍丽娜, 薛乐仁, 戴钰俊, 赵新宇, 王世行, 王威. 基于扩张重参数化和空洞卷积架构的步态识别方法[J]. 计算机应用, 2025, 45(4): 1285-1292.
Lina HUO, Leren XUE, Yujun DAI, Xinyu ZHAO, Shihang WANG, Wei WANG. Gait recognition method based on dilated reparameterization and atrous convolution architecture[J]. Journal of Computer Applications, 2025, 45(4): 1285-1292.
方法 | Rank-1 | Rank-5 | mAP | mINP |
---|---|---|---|---|
GaitPart[ | 29.9 | 50.6 | 23.3 | 13.2 |
GaitSet[ | 42.6 | 63.1 | 33.7 | 19.7 |
GLN[ | 42.2 | 64.5 | 33.1 | 19.6 |
CSTL[ | 12.2 | 21.7 | 6.4 | 3.3 |
GaitGL[ | 23.5 | 38.5 | 16.4 | 9.2 |
SMPLGait[ | 53.2 | 71.0 | 42.4 | 26.0 |
GaitTAKE[ | 53.1 | 71.9 | 43.3 | 27.1 |
GaitRef[ | 54.1 | 71.3 | 42.7 | 26.2 |
GaitGCI[ | 57.2 | 74.5 | 45.0 | 27.6 |
HSTL[ | 61.3 | 76.3 | 55.5 | 34.8 |
GaitBase[ | 64.6 | 79.8 | 53.4 | 31.4 |
DilatedGait | 73.6 | 87.4 | 65.0 | 45.6 |
表1 在Gait3D数据集上不同方法的性能对比 (%)
Tab. 1 Performance comparison of different methods on Gait3D dataset
方法 | Rank-1 | Rank-5 | mAP | mINP |
---|---|---|---|---|
GaitPart[ | 29.9 | 50.6 | 23.3 | 13.2 |
GaitSet[ | 42.6 | 63.1 | 33.7 | 19.7 |
GLN[ | 42.2 | 64.5 | 33.1 | 19.6 |
CSTL[ | 12.2 | 21.7 | 6.4 | 3.3 |
GaitGL[ | 23.5 | 38.5 | 16.4 | 9.2 |
SMPLGait[ | 53.2 | 71.0 | 42.4 | 26.0 |
GaitTAKE[ | 53.1 | 71.9 | 43.3 | 27.1 |
GaitRef[ | 54.1 | 71.3 | 42.7 | 26.2 |
GaitGCI[ | 57.2 | 74.5 | 45.0 | 27.6 |
HSTL[ | 61.3 | 76.3 | 55.5 | 34.8 |
GaitBase[ | 64.6 | 79.8 | 53.4 | 31.4 |
DilatedGait | 73.6 | 87.4 | 65.0 | 45.6 |
方法 | Rank-1 | Rank-5 | Rank-10 | Rank-20 |
---|---|---|---|---|
PoseGait[ | 0.2 | 1.1 | 2.2 | 4.3 |
GaitGraph[ | 1.3 | 3.5 | 5.1 | 7.5 |
GaitPart[ | 44.0 | 60.7 | 67.4 | 73.5 |
GaitSet[ | 46.3 | 63.6 | 70.3 | 76.8 |
CSTL[ | 50.6 | 65.9 | 71.9 | 76.9 |
GaitGL[ | 51.4 | 67.5 | 72.8 | 77.3 |
GaitTAKE[ | 51.3 | 69.4 | 75.5 | 80.4 |
GaitMix[ | 52.4 | 67.4 | 72.9 | 77.2 |
GaitRef[ | 53.0 | 67.9 | 73.0 | 77.5 |
TransGait[ | 56.3 | 72.7 | 78.1 | 82.5 |
GaitBase[ | 60.1 | 75.5 | 80.4 | 84.1 |
DilatedGait | 71.7 | 84.3 | 88.1 | 90.8 |
表2 在GREW数据集上不同方法的性能对比 (%)
Tab. 2 Performance comparison of different methods on GREW dataset
方法 | Rank-1 | Rank-5 | Rank-10 | Rank-20 |
---|---|---|---|---|
PoseGait[ | 0.2 | 1.1 | 2.2 | 4.3 |
GaitGraph[ | 1.3 | 3.5 | 5.1 | 7.5 |
GaitPart[ | 44.0 | 60.7 | 67.4 | 73.5 |
GaitSet[ | 46.3 | 63.6 | 70.3 | 76.8 |
CSTL[ | 50.6 | 65.9 | 71.9 | 76.9 |
GaitGL[ | 51.4 | 67.5 | 72.8 | 77.3 |
GaitTAKE[ | 51.3 | 69.4 | 75.5 | 80.4 |
GaitMix[ | 52.4 | 67.4 | 72.9 | 77.2 |
GaitRef[ | 53.0 | 67.9 | 73.0 | 77.5 |
TransGait[ | 56.3 | 72.7 | 78.1 | 82.5 |
GaitBase[ | 60.1 | 75.5 | 80.4 | 84.1 |
DilatedGait | 71.7 | 84.3 | 88.1 | 90.8 |
编号 | 方案 | Rank-1 | Rank-5 | mAP | mINP |
---|---|---|---|---|---|
Baseline(对照组) | 68.2 | 84.6 | 60.4 | 40.0 | |
① | +扩张卷积(3×3, r=2) +常规卷积(3×3) | 69.8 | 84.8 | 62.0 | 41.5 |
② | +扩张卷积(3×3, r=2) +常规卷积(5×5) | 70.2 | 85.4 | 62.8 | 42.7 |
③ | +扩张卷积(3×3, r=3) +常规卷积(3×3) +常规卷积(7×7) | 70.3 | 85.2 | 62.9 | 42.6 |
④ | +扩张卷积(3×3, r=2) +常规卷积(3×3) +常规卷积(5×5) | 70.8 | 85.6 | 63.5 | 43.2 |
表3 DRM的消融实验结果 (%)
Tab. 3 Results of ablation experiments of DRM
编号 | 方案 | Rank-1 | Rank-5 | mAP | mINP |
---|---|---|---|---|---|
Baseline(对照组) | 68.2 | 84.6 | 60.4 | 40.0 | |
① | +扩张卷积(3×3, r=2) +常规卷积(3×3) | 69.8 | 84.8 | 62.0 | 41.5 |
② | +扩张卷积(3×3, r=2) +常规卷积(5×5) | 70.2 | 85.4 | 62.8 | 42.7 |
③ | +扩张卷积(3×3, r=3) +常规卷积(3×3) +常规卷积(7×7) | 70.3 | 85.2 | 62.9 | 42.6 |
④ | +扩张卷积(3×3, r=2) +常规卷积(3×3) +常规卷积(5×5) | 70.8 | 85.6 | 63.5 | 43.2 |
实验方法 | Gait3D | GREW | ||||||
---|---|---|---|---|---|---|---|---|
Rank-1 | Rank-5 | mAP | mINP | Rank-1 | Rank-5 | Rank-10 | Rank-20 | |
Baseline(对照组) | 68.2 | 84.6 | 60.4 | 40.0 | 68.6 | 82.0 | 86.1 | 88.2 |
+DRM | 70.8 | 85.6 | 63.5 | 43.2 | 69.8 | 83.1 | 86.5 | 88.9 |
+残差空洞卷积块 | 72.7 | 87.2 | 64.7 | 45.2 | 71.0 | 84.1 | 87.8 | 90.5 |
DilatedGait | 73.6 | 87.4 | 65.0 | 45.6 | 71.7 | 84.3 | 88.1 | 90.8 |
表4 方法组合在Gait3D与GREW数据集上的准确率对比 (%)
Tab. 4 Comparison of accuracy of method combinations on Gait3D and GREW datasets
实验方法 | Gait3D | GREW | ||||||
---|---|---|---|---|---|---|---|---|
Rank-1 | Rank-5 | mAP | mINP | Rank-1 | Rank-5 | Rank-10 | Rank-20 | |
Baseline(对照组) | 68.2 | 84.6 | 60.4 | 40.0 | 68.6 | 82.0 | 86.1 | 88.2 |
+DRM | 70.8 | 85.6 | 63.5 | 43.2 | 69.8 | 83.1 | 86.5 | 88.9 |
+残差空洞卷积块 | 72.7 | 87.2 | 64.7 | 45.2 | 71.0 | 84.1 | 87.8 | 90.5 |
DilatedGait | 73.6 | 87.4 | 65.0 | 45.6 | 71.7 | 84.3 | 88.1 | 90.8 |
1 | 刘瑞华,郝子赫,邹洋杨. 基于多层级精细特征融合的步态识别算法[J]. 计算机应用, 2024, 44(7):2250-2257. |
LIU R H, HAO Z H, ZOU Y Y. Gait recognition algorithm based on multi-layer refined feature fusion[J]. Journal of Computer Applications, 2024, 44(7):2250-2257. | |
2 | SHEN C, YU S, WANG J, et al. A comprehensive survey on deep gait recognition: algorithms, datasets and challenges[EB/OL]. [2024-04-03].. |
3 | LIAO R, YU S, AN W, et al. A model-based gait recognition method with body pose and human prior knowledge[J]. Pattern Recognition, 2020, 98: No.107069. |
4 | TEEPE T, KHAN A, GILG J, et al. GaitGraph: graph convolutional network for skeleton-based gait recognition[C]// Proceedings of the 2021 IEEE International Conference on Image Processing. Piscataway: IEEE, 2021: 2314-2318. |
5 | ZHENG J, LIU X, LIU W, et al. Gait recognition in the wild with dense 3D representations and a benchmark[C]// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 20196-20205. |
6 | HUANG X, ZHU D, WANG H, et al. Context-sensitive temporal feature learning for gait recognition[C]// Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 12889-12898. |
7 | HSU H M, WANG Y, YANG C Y, et al. GaitTAKE: gait recognition by temporal attention and keypoint-guided embedding[C]// Proceedings of the 2022 IEEE International Conference on Image Processing. Piscataway: IEEE, 2022: 2546-2550. |
8 | WANG L, LIU B, LIANG F, et al. Hierarchical spatio-temporal representation learning for gait recognition[C]// Proceedings of the 2023 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2023: 19582-19592. |
9 | DOU H, ZHANG P, SU W, et al. GaitGCI: generative counterfactual intervention for gait recognition[C]// Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 5578-5588. |
10 | CHAO H, WANG K, HE Y, et al. GaitSet: cross-view gait recognition through utilizing gait as a deep set[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(7): 3467-3478. |
11 | FAN C, PENG Y, CAO C, et al. GaitPart: temporal part-based model for gait recognition[C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 14213-14221. |
12 | LIN B, ZHANG S, YU X. Gait recognition via effective global-local feature representation and local temporal aggregation[C]// Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 14628-14636. |
13 | HOU S, CAO C, LIU X, et al. Gait lateral network: learning discriminative and compact representations for gait recognition[C]// Proceedings of the 2020 European Conference on Computer Vision, LNCS 12354. Cham: Springer, 2020: 382-398. |
14 | ZHU H, ZHENG W, ZHENG Z, et al. GaitRef: gait recognition with refined sequential skeletons[C]// Proceedings of the 2023 IEEE International Joint Conference on Biometrics. Piscataway: IEEE, 2023: 1-10. |
15 | LI G, GUO L, ZHANG R, et al. TransGait: multimodal-based gait recognition with set transformer[J]. Applied Intelligence, 2023, 53(2): 1535-1547. |
16 | FAN C, HOU S, HUANG Y, et al. Exploring deep models for practical gait recognition[EB/OL]. [2024-04-03].. |
17 | FAN C, LIANG J, SHEN C, et al. OpenGait: revisiting gait recognition toward better practicality[C]// Proceedings of the 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2023: 9707-9716. |
18 | HE K, ZHANG X, REN S, 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. |
19 | ZHANG Y, MAZUROWSKI M A. Convolutional neural networks rarely learn shape for semantic segmentation[J]. Pattern Recognition, 2024, 146: No.110018. |
20 | LUO W, LI Y, URTASUN R, et al. Understanding the effective receptive field in deep convolutional neural networks[C]// Proceedings of the 30th International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2016: 4905-4913. |
21 | YU F, KOLTUN V, FUNKHOUSER T. Dilated residual networks[C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 636-644. |
22 | DING X, GUO Y, DING G, et al. ACNet: strengthening the kernel skeletons for powerful CNN via asymmetric convolution blocks[C]// Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2019: 1911-1920. |
23 | 邓帆,曾渊,刘博文,等. 基于Transformer时间特征聚合的步态识别模型[J]. 计算机应用, 2023, 43(S1): 15-18. |
DENG F, ZENG Y, LIU B W, et al. Gait recognition model based on temporal feature aggregation with Transformer[J]. Journal of Computer Applications, 2023, 43(S1): 15-18. | |
24 | LUO H, GU Y, LIAO X, et al. Bag of tricks and a strong baseline for deep person re-identification[C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Piscataway: IEEE, 2019: 1487-1495. |
25 | GLOROT X, BORDES A, BENGIO Y. Deep sparse rectifier neural networks[C]// Proceedings of the 14th International Conference on Artificial Intelligence and Statistics. New York: JMLR.org, 2011: 315-323. |
26 | CHOLLET F. Xception: deep learning with depthwise separable convolutions[C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 1800-1807. |
27 | LIU Z, LIN Y, CAO Y, et al. Swin Transformer: hierarchical vision Transformer using shifted windows[C]// Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 9992-10002. |
28 | DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An image is worth 16x16 words: Transformers for image recognition at scale[EB/OL]. [2024-04-03].. |
29 | DING X, ZHANG X, HAN J, et al. Scaling up your kernels to 31×31: revisiting large kernel design in CNNs[C]// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 11953-11965. |
30 | 马玉祥,代雪晶. 基于视频残差神经网络的深度步态识别[J]. 计算机系统应用, 2024, 33(4): 279-287. |
MA Y X, DAI X J. Deep gait recognition based on video residual neural network[J]. Computer Systems and Applications, 2024, 33(4): 279-287. | |
31 | SCHROFF F, KALENICHENKO D, PHILBIN J. FaceNet: a unified embedding for face recognition and clustering[C]// Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2015: 815-823. |
32 | ZHU Z, GUO X, YANG T, et al. Gait recognition in the wild: a benchmark[C]// Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 14769-14779. |
33 | YE M, SHEN J, LIN G, et al. Deep learning for person re-identification: a survey and outlook[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(6): 2872-2893. |
[1] | 赵志强, 马培红, 黑新宏. 基于双重注意力机制的人群计数方法[J]. 《计算机应用》唯一官方网站, 2024, 44(9): 2886-2892. |
[2] | 刘瑞华, 郝子赫, 邹洋杨. 基于多层级精细特征融合的步态识别算法[J]. 《计算机应用》唯一官方网站, 2024, 44(7): 2250-2257. |
[3] | 王晓路, 千王菲. 基于双支路卷积网络的步态识别方法[J]. 《计算机应用》唯一官方网站, 2024, 44(6): 1965-1971. |
[4] | 封筠, 毕健康, 霍一儒, 李家宽. 轻量化沥青路面裂缝图像分割网络PIPNet[J]. 《计算机应用》唯一官方网站, 2024, 44(5): 1520-1526. |
[5] | 蒋占军, 吴佰靖, 马龙, 廉敬. 多尺度特征和极化自注意力的Faster-RCNN水漂垃圾识别[J]. 《计算机应用》唯一官方网站, 2024, 44(3): 938-944. |
[6] | 王林, 刘景亮, 王无为. 基于空洞卷积融合Transformer的无人机图像小目标检测方法[J]. 《计算机应用》唯一官方网站, 2024, 44(11): 3595-3602. |
[7] | 梁美佳, 刘昕武, 胡晓鹏. 基于改进YOLOv3的列车运行环境图像小目标检测算法[J]. 《计算机应用》唯一官方网站, 2023, 43(8): 2611-2618. |
[8] | 刘辉, 张琳玉, 王复港, 何如瑾. 基于注意力机制和上下文信息的目标检测算法[J]. 《计算机应用》唯一官方网站, 2023, 43(5): 1557-1564. |
[9] | 刘今越, 李慧宇, 贾晓辉, 李佳蕊. 基于人体模型约束的步态动态识别方法[J]. 《计算机应用》唯一官方网站, 2023, 43(3): 972-977. |
[10] | 李佳东, 张丹普, 范亚琼, 杨剑锋. 基于改进YOLOv5的轻量级船舶目标检测算法[J]. 《计算机应用》唯一官方网站, 2023, 43(3): 923-929. |
[11] | 张志昂, 廖光忠. 基于U-Net的多尺度特征增强视网膜血管分割算法[J]. 《计算机应用》唯一官方网站, 2023, 43(10): 3275-3281. |
[12] | 余晓鹏, 何儒汉, 黄晋, 张俊杰, 胡新荣. 基于改进Inception结构的知识图谱嵌入模型[J]. 《计算机应用》唯一官方网站, 2022, 42(4): 1065-1071. |
[13] | 刘朋伟, 高媛, 秦品乐, 殷喆, 王丽芳. 基于多感受野的生成对抗网络医学MRI影像超分辨率重建[J]. 《计算机应用》唯一官方网站, 2022, 42(3): 938-945. |
[14] | 许慧青, 陈斌, 王敬飞, 陈志毅, 覃健. 基于卷积神经网络的细长路面病害检测方法[J]. 《计算机应用》唯一官方网站, 2022, 42(1): 265-272. |
[15] | 冯兴杰, 张天泽. 基于分组卷积进行特征融合的全景分割算法[J]. 计算机应用, 2021, 41(7): 2054-2061. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||