《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (2): 583-588.DOI: 10.11772/j.issn.1001-9081.2021122075
• 多媒体计算与计算机仿真 • 上一篇
收稿日期:
2021-12-09
修回日期:
2022-02-20
接受日期:
2022-02-23
发布日期:
2023-02-08
出版日期:
2023-02-10
通讯作者:
刘翠响
作者简介:
苏亚婷(1995—),女,河北石家庄人,硕士研究生,主要研究方向:信息感知、机器学习;
基金资助:
Received:
2021-12-09
Revised:
2022-02-20
Accepted:
2022-02-23
Online:
2023-02-08
Published:
2023-02-10
Contact:
Cuixiang LIU
About author:
SU Yating, born in 1995, M. S. candidate. Her research interests include information perception, machine learning.
Supported by:
摘要:
针对单目图像重建人体时出现的头部姿态翻转和图像特征间隐式空间线索缺失的问题,提出了一种基于高分辨率网络(HRNet)和图卷积网络(GCN)的三维人体重建模型。首先利用HRNet和残差块作为主干网络从原始图像中提取丰富的人体特征信息,然后使用GCN来捕获特征之间隐式的空间线索以获得空间精确的特征表示,最后使用此特征来预测多人线性蒙皮模型(SMPL)的参数以得到更加准确的重建结果;同时为了有效解决人体头部姿态翻转的问题,对SMPL的关节点重新进行了定义,在原有关节的基础上增加对头部关节点的定义。实验结果表明,所提模型能够准确地重建出三维人体,在2D数据集LSP上的重建准确率达到了92.41%,在3D数据集MPI-INF-3DHP上的关节误差和重建误差也大幅降低,平均误差仅分别为97.73 mm和64.63 mm,验证了所提模型在人体重建领域的有效性。
中图分类号:
苏亚婷, 刘翠响. 基于高分辨率网络和图卷积网络的三维人体重建模型[J]. 计算机应用, 2023, 43(2): 583-588.
Yating SU, Cuixiang LIU. Three-dimensional human reconstruction model based on high-resolution net and graph convolutional network[J]. Journal of Computer Applications, 2023, 43(2): 583-588.
模型 | F1 | 准确率 | 模型 | F1 | 准确率 |
---|---|---|---|---|---|
SMPLify | 84.90 | 90.56 | CMR | 87.10 | 91.55 |
HMR | 86.95 | 91.02 | 本文模型 | 88.03 | 92.41 |
表1 重建性能比较 (%)
Tab. 1 Reconstruction performance comparison
模型 | F1 | 准确率 | 模型 | F1 | 准确率 |
---|---|---|---|---|---|
SMPLify | 84.90 | 90.56 | CMR | 87.10 | 91.55 |
HMR | 86.95 | 91.02 | 本文模型 | 88.03 | 92.41 |
视频帧 | SMPLify | HMR | CMR | 本文 |
---|---|---|---|---|
平均 | 943.57 | 235.73 | 181.80 | 97.73 |
TS1 | 844.13 | 187.09 | 145.70 | 63.36 |
TS2 | 897.08 | 283.63 | 172.57 | 89.76 |
TS3 | 1 059.01 | 251.29 | 160.07 | 91.96 |
TS4 | 974.92 | 265.72 | 233.98 | 106.48 |
TS5 | 856.23 | 172.19 | 208.39 | 116.07 |
TS6 | 1 030.02 | 254.45 | 170.09 | 118.73 |
表2 MPJPE误差结果 (mm)
Tab. 2 MPJPE error results
视频帧 | SMPLify | HMR | CMR | 本文 |
---|---|---|---|---|
平均 | 943.57 | 235.73 | 181.80 | 97.73 |
TS1 | 844.13 | 187.09 | 145.70 | 63.36 |
TS2 | 897.08 | 283.63 | 172.57 | 89.76 |
TS3 | 1 059.01 | 251.29 | 160.07 | 91.96 |
TS4 | 974.92 | 265.72 | 233.98 | 106.48 |
TS5 | 856.23 | 172.19 | 208.39 | 116.07 |
TS6 | 1 030.02 | 254.45 | 170.09 | 118.73 |
视频帧 | SMPLify | HMR | CMR | 本文 |
---|---|---|---|---|
平均 | 138.85 | 130.63 | 97.38 | 64.63 |
TS1 | 171.14 | 102.07 | 75.29 | 41.72 |
TS2 | 145.51 | 132.44 | 112.70 | 60.29 |
TS3 | 123.27 | 142.19 | 91.94 | 58.60 |
TS4 | 135.35 | 152.72 | 110.51 | 66.00 |
TS5 | 138.76 | 108.19 | 85.66 | 73.86 |
TS6 | 119.09 | 146.15 | 108.15 | 87.31 |
表3 重建误差结果 (mm)
Tab. 3 Reconstruction error results
视频帧 | SMPLify | HMR | CMR | 本文 |
---|---|---|---|---|
平均 | 138.85 | 130.63 | 97.38 | 64.63 |
TS1 | 171.14 | 102.07 | 75.29 | 41.72 |
TS2 | 145.51 | 132.44 | 112.70 | 60.29 |
TS3 | 123.27 | 142.19 | 91.94 | 58.60 |
TS4 | 135.35 | 152.72 | 110.51 | 66.00 |
TS5 | 138.76 | 108.19 | 85.66 | 73.86 |
TS6 | 119.09 | 146.15 | 108.15 | 87.31 |
层数N | 重建 误差/mm | MPJPE/mm | 层数N | 重建 误差/mm | MPJPE/mm |
---|---|---|---|---|---|
0 | 181.52 | 447.78 | 3 | 105.78 | 180.05 |
1 | 183.90 | 259.01 | 4 | 81.84 | 117.69 |
2 | 122.02 | 224.78 | 5 | 55.61 | 88.60 |
表4 MPI-INF-3DPH数据集上的消融实验
Tab. 4 Ablation experiment on MPI-INF-3DPH dataset
层数N | 重建 误差/mm | MPJPE/mm | 层数N | 重建 误差/mm | MPJPE/mm |
---|---|---|---|---|---|
0 | 181.52 | 447.78 | 3 | 105.78 | 180.05 |
1 | 183.90 | 259.01 | 4 | 81.84 | 117.69 |
2 | 122.02 | 224.78 | 5 | 55.61 | 88.60 |
头部姿态约束 | 重建误差 | ||
---|---|---|---|
TS2 | TS4 | TS6 | |
无 | 112.49 | 114.12 | 136.36 |
有 | 63.48 | 66.88 | 89.95 |
表5 头部关节对重建误差的影响 (mm)
Tab. 5 Influence of head joints on reconstruction error
头部姿态约束 | 重建误差 | ||
---|---|---|---|
TS2 | TS4 | TS6 | |
无 | 112.49 | 114.12 | 136.36 |
有 | 63.48 | 66.88 | 89.95 |
1 | 杨继魁. 基于Kinect单次拍摄数据准确估计人体全身体型与姿态的研究[D]. 合肥:安徽大学, 2019:10-16. |
YANG J K. Accurately estimating the whole body shape and pose of human body based on Kinect single shot data[D]. Hefei: Anhui University, 2019: 10-16. | |
2 | LOPER M, MAHMOOD N, ROMERO J, et al. SMPL: a skinned multi-person linear model[J]. ACM Transactions on Graphics, 2015, 34(6): No.248. 10.1145/2816795.2818013 |
3 | SUN K, XIAO B, LIU D, et al. Deep high-resolution representation learning for human pose estimation[C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 5686-5696. 10.1109/cvpr.2019.00584 |
4 | 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 |
5 | 张亚凤,刘翠响,马杰,等. 基于多特征点匹配的三维人体姿态重建[J]. 激光与光电子学进展, 2022, 59(16):325-332. 10.3788/lop202259.1615003 |
ZHANG Y F, LIU C X, MA J, et al. Three-dimensional human pose reconstruction based on multifeature point matching[J]. Laser and Optoelectronics Progress, 2022, 59(16):325-332. 10.3788/lop202259.1615003 | |
6 | ANGUELOV D, SRINIVASAN P, KOLLER D, et al. SCAPE: shape completion and animation of people[J]. ACM Transactions on Graphics, 2005, 24(3): 408-416. 10.1145/1073204.1073207 |
7 | GUAN P, WEISS A, BĂLAN A O, et al. Estimating human shape and pose from a single image[C]// Proceedings of the IEEE 12th International Conference on Computer Vision. Piscataway: IEEE, 2009: 1381-1388. 10.1109/iccv.2009.5459300 |
8 | BĂLAN A O, SIGAL L, BLACK M J, et al. Detailed human shape and pose from images[C]// Proceedings of the 2007 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2007: 1-8. 10.1109/cvpr.2007.383340 |
9 | BOBO F, KANAZAWA A, LASSNER C, et al. Keep it SMPL: automatic estimation of 3D human pose and shape from a single image[C]// Proceedings of the 2016 European Conference on Computer Vision, LNCS 9909. Cham: Springer, 2016: 561-578. |
10 | HUANG Y H, BOGO F, LASSNER C, et al. Towards accurate marker-less human shape and pose estimation over time[C]// Proceedings of the 2017 International Conference on 3D Vision. Piscataway: IEEE, 2017: 421-430. 10.1109/3dv.2017.00055 |
11 | LASSNER C, ROMERO J, KIEFEL M, et al. Unite the people: closing the loop between 3D and 2D human representations[C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 4704-4713. 10.1109/cvpr.2017.500 |
12 | ZANFIR A, MARINOIU E, SMINCHISESCU C. Monocular 3D pose and shape estimation of multiple people in natural scenes —the importance of multiple scene constraints[C]// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 2148-2157. 10.1109/cvpr.2018.00229 |
13 | DIBRA E, JAIN H, ÖZTIRELI C, et al. HS-Nets: estimating human body shape from silhouettes with convolutional neural networks[C]// Proceedings of the 4th International Conference on 3D Vision. Piscataway: IEEE, 2016: 108-117. 10.1109/3dv.2016.19 |
14 | TAN J K V, BUDVYTIS I, CIPOLLA R. Indirect deep structured learning for 3D human body shape and pose prediction[C]// Proceedings of the 2017 British Machine Vision Conference Durham: BMVA Press, 2017: No.722. 10.5244/c.31.15 |
15 | TUNG H Y F, TUNG H W, YUMER E, et al. Self-supervised learning of motion capture[C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc., 2017:5242-5252. |
16 | KANAZAWA A, BLACK M J, JACOBS D W, et al. End-to-end recovery of human shape and pose[C]// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 7122-7131. 10.1109/cvpr.2018.00744 |
17 | KOLOTOUROS N, PAVLAKOS G, BLACK M J, et al. Learning to reconstruct 3D human pose and shape via model-fitting in the loop[C]// Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2019: 2252-2261. 10.1109/iccv.2019.00234 |
18 | ZHANG T S, HUANG B Z, WANG Y G. Object-occluded human shape and pose estimation from a single color image[C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 7374-7383. 10.1109/cvpr42600.2020.00740 |
19 | LI Z G, OSKARSSON M, HEYDEN A. 3D human pose and shape estimation through collaborative learning and multi-view model-fitting[C]// Proceedings of the 2021 IEEE/CVF Winter Conference on Applications of Computer Vision. Piscataway: IEEE, 2021: 1887-1896. 10.1109/wacv48630.2021.00193 |
20 | KOLOTOUROS N, PAVLAKOS G, DANIILIDIS K. Convolutional mesh regression for single-image human shape reconstruction[C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 4496-4505. 10.1109/cvpr.2019.00463 |
21 | XIE H Y, ZHONG Y Q, YU Z C, et al. Non-parametric anthropometric graph convolutional network for virtual mannequin reconstruction[J]. IEEE Access, 2020, 8: 3539-3550. 10.1109/access.2019.2962833 |
22 | ZHANG S Z, XIAO N F. Detailed 3D human body reconstruction from a single image based on mesh deformation[J]. IEEE Access, 2021, 9: 8595-8603. 10.1109/access.2021.3049548 |
23 | CHENG K L, TONG R F, TANG M, et al. Parametric human body reconstruction based on sparse key points[J]. IEEE Transactions on Visualization and Computer Graphics, 2016, 22(11): 2467-2479. 10.1109/tvcg.2015.2511751 |
24 | BOGO F, ROMERO J, LOPER M, et al. FAUST: dataset and evaluation for 3D mesh registration[C]// Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2014: 3794-3801. 10.1109/cvpr.2014.491 |
25 | von MARCARD T, HENSCHEL R, BLACK M J, et al. Recovering accurate 3D human pose in the wild using IMUs and a moving camera[C]// Proceedings of the 2018 European Conference on Computer Vision, LNCS 11214. Cham: Springer, 2018: 614-631. |
26 | LIN T Y, MAIRE M, BELONGIE S, et al. Microsoft COCO: common objects in context[C]// Proceedings of the 2014 European Conference on Computer Vision, LNCS 8693. Cham: Springer, 2014: 740-755. |
27 | ANDRILUKA M, PISHCHULIN L, GEHLER P, et al. 2D human pose estimation: new benchmark and state of the art analysis[C]// Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2014: 3686-3693. 10.1109/cvpr.2014.471 |
28 | JOHNSON S, EVERINGHAM M. Clustered pose and nonlinear appearance models for human pose estimation[C]// Proceedings of the 2010 British Machine Vision Conference. Durham: BMVA Press, 2010: No.12. 10.5244/c.24.12 |
29 | MEHTA D, RHODIN H, CASAS D, et al. Monocular 3D human pose estimation in the wild using improved CNN supervision[C]// Proceedings of the 2017 International Conference on 3D Vision. Piscataway: IEEE, 2017: 506-516. 10.1109/3dv.2017.00064 |
[1] | 袁立宁, 刘钊. 基于One-Shot聚合自编码器的图表示学习[J]. 《计算机应用》唯一官方网站, 2023, 43(1): 8-14. |
[2] | 李坤, 侯庆. 基于注意力机制的轻量型人体姿态估计[J]. 《计算机应用》唯一官方网站, 2022, 42(8): 2407-2414. |
[3] | 刘博, 卿粼波, 王正勇, 刘美, 姜雪. 基于分块注意力机制和交互位置关系的群组活动识别[J]. 《计算机应用》唯一官方网站, 2022, 42(7): 2052-2057. |
[4] | 左亚尧, 陈皓宇, 陈致然, 洪嘉伟, 陈坤. 融合多语义特征的命名实体识别方法[J]. 《计算机应用》唯一官方网站, 2022, 42(7): 2001-2008. |
[5] | 文敏, 王荣存, 姜淑娟. 基于关系图卷积网络的源代码漏洞检测[J]. 《计算机应用》唯一官方网站, 2022, 42(6): 1814-1821. |
[6] | 孙琪翔, 何宁, 张敬尊, 宏晨. 基于非局部高分辨率网络的轻量化人体姿态估计方法[J]. 《计算机应用》唯一官方网站, 2022, 42(5): 1398-1406. |
[7] | 张晔, 刘蓉, 刘明, 陈明. 基于多通道注意力机制的图像超分辨率重建网络[J]. 《计算机应用》唯一官方网站, 2022, 42(5): 1563-1569. |
[8] | 杜雨奇, 郑津, 王杨, 黄诚, 李平. 基于图卷积网络的文本分割模型[J]. 《计算机应用》唯一官方网站, 2022, 42(12): 3692-3699. |
[9] | 张斌, 王莉, 杨延杰. 联合立场的过程跟踪式多任务谣言验证模型[J]. 《计算机应用》唯一官方网站, 2022, 42(11): 3371-3378. |
[10] | 李卓然, 冶忠林, 赵海兴, 林晶晶. 基于混合特征建模的图卷积网络方法[J]. 《计算机应用》唯一官方网站, 2022, 42(11): 3354-3363. |
[11] | 谢斌红, 李书宁, 张英俊. 基于层次结构感知的细粒度实体分类方法[J]. 《计算机应用》唯一官方网站, 2022, 42(10): 3003-3010. |
[12] | 刘长红, 曾胜, 张斌, 陈勇. 基于语义关系图的跨模态张量融合网络的图像文本检索[J]. 《计算机应用》唯一官方网站, 2022, 42(10): 3018-3024. |
[13] | 张继杰, 杨艳, 刘勇. 利用初始残差和解耦操作的自适应深层图卷积[J]. 《计算机应用》唯一官方网站, 2022, 42(1): 9-15. |
[14] | 李扬, 吴安彪, 袁野, 赵琳琳, 王国仁. 基于节点相似度的无监督属性图嵌入模型[J]. 《计算机应用》唯一官方网站, 2022, 42(1): 1-8. |
[15] | 马栋林, 马司周, 王伟杰. 基于图卷积网络和门控循环单元的多站点气温预测模型[J]. 《计算机应用》唯一官方网站, 2022, 42(1): 287-293. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||