Attention-Guided Huber Loss for Head Pose Estimation Based on Improved Capsule Network
Head pose estimation is an important technology for analyzing human behavior and has been widely researched and applied in areas such as human–computer interaction and fatigue detection. However, traditional head pose estimation networks suffer from the problem of easily losing spatial structure inf...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-07-01
|
| Series: | Entropy |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1099-4300/25/7/1024 |
| _version_ | 1797589362215485440 |
|---|---|
| author | Runhao Zhong Li He Hongwei Wang Liang Yuan Kexin Li Zhening Liu |
| author_facet | Runhao Zhong Li He Hongwei Wang Liang Yuan Kexin Li Zhening Liu |
| author_sort | Runhao Zhong |
| collection | DOAJ |
| description | Head pose estimation is an important technology for analyzing human behavior and has been widely researched and applied in areas such as human–computer interaction and fatigue detection. However, traditional head pose estimation networks suffer from the problem of easily losing spatial structure information, particularly in complex scenarios where occlusions and multiple object detections are common, resulting in low accuracy. To address the above issues, we propose a head pose estimation model based on the residual network and capsule network. Firstly, a deep residual network is used to extract features from three stages, capturing spatial structure information at different levels, and a global attention block is employed to enhance the spatial weight of feature extraction. To effectively avoid the loss of spatial structure information, the features are encoded and transmitted to the output using an improved capsule network, which is enhanced in its generalization ability through self-attention routing mechanisms. To enhance the robustness of the model, we optimize Huber loss, which is first used in head pose estimation. Finally, experiments are conducted on three popular public datasets, 300W-LP, AFLW2000, and BIWI. The results demonstrate that the proposed method achieves state-of-the-art results, particularly in scenarios with occlusions. |
| first_indexed | 2024-03-11T01:06:36Z |
| format | Article |
| id | doaj.art-44a2d1cfa6ff4968b056d9ffcdd92090 |
| institution | Directory Open Access Journal |
| issn | 1099-4300 |
| language | English |
| last_indexed | 2024-03-11T01:06:36Z |
| publishDate | 2023-07-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Entropy |
| spelling | doaj.art-44a2d1cfa6ff4968b056d9ffcdd920902023-11-18T19:13:37ZengMDPI AGEntropy1099-43002023-07-01257102410.3390/e25071024Attention-Guided Huber Loss for Head Pose Estimation Based on Improved Capsule NetworkRunhao Zhong0Li He1Hongwei Wang2Liang Yuan3Kexin Li4Zhening Liu5School of Mechanical Engineering, Xinjiang University, Urumqi 830046, ChinaSchool of Mechanical Engineering, Xinjiang University, Urumqi 830046, ChinaSchool of Mechanical Engineering, Xinjiang University, Urumqi 830046, ChinaSchool of Mechanical Engineering, Xinjiang University, Urumqi 830046, ChinaSchool of Mechanical Engineering, Xinjiang University, Urumqi 830046, ChinaSchool of Mechanical Engineering, Xinjiang University, Urumqi 830046, ChinaHead pose estimation is an important technology for analyzing human behavior and has been widely researched and applied in areas such as human–computer interaction and fatigue detection. However, traditional head pose estimation networks suffer from the problem of easily losing spatial structure information, particularly in complex scenarios where occlusions and multiple object detections are common, resulting in low accuracy. To address the above issues, we propose a head pose estimation model based on the residual network and capsule network. Firstly, a deep residual network is used to extract features from three stages, capturing spatial structure information at different levels, and a global attention block is employed to enhance the spatial weight of feature extraction. To effectively avoid the loss of spatial structure information, the features are encoded and transmitted to the output using an improved capsule network, which is enhanced in its generalization ability through self-attention routing mechanisms. To enhance the robustness of the model, we optimize Huber loss, which is first used in head pose estimation. Finally, experiments are conducted on three popular public datasets, 300W-LP, AFLW2000, and BIWI. The results demonstrate that the proposed method achieves state-of-the-art results, particularly in scenarios with occlusions.https://www.mdpi.com/1099-4300/25/7/1024head pose estimationglobal attention blockself-attention routingcapsule network |
| spellingShingle | Runhao Zhong Li He Hongwei Wang Liang Yuan Kexin Li Zhening Liu Attention-Guided Huber Loss for Head Pose Estimation Based on Improved Capsule Network Entropy head pose estimation global attention block self-attention routing capsule network |
| title | Attention-Guided Huber Loss for Head Pose Estimation Based on Improved Capsule Network |
| title_full | Attention-Guided Huber Loss for Head Pose Estimation Based on Improved Capsule Network |
| title_fullStr | Attention-Guided Huber Loss for Head Pose Estimation Based on Improved Capsule Network |
| title_full_unstemmed | Attention-Guided Huber Loss for Head Pose Estimation Based on Improved Capsule Network |
| title_short | Attention-Guided Huber Loss for Head Pose Estimation Based on Improved Capsule Network |
| title_sort | attention guided huber loss for head pose estimation based on improved capsule network |
| topic | head pose estimation global attention block self-attention routing capsule network |
| url | https://www.mdpi.com/1099-4300/25/7/1024 |
| work_keys_str_mv | AT runhaozhong attentionguidedhuberlossforheadposeestimationbasedonimprovedcapsulenetwork AT lihe attentionguidedhuberlossforheadposeestimationbasedonimprovedcapsulenetwork AT hongweiwang attentionguidedhuberlossforheadposeestimationbasedonimprovedcapsulenetwork AT liangyuan attentionguidedhuberlossforheadposeestimationbasedonimprovedcapsulenetwork AT kexinli attentionguidedhuberlossforheadposeestimationbasedonimprovedcapsulenetwork AT zheningliu attentionguidedhuberlossforheadposeestimationbasedonimprovedcapsulenetwork |