Human Pose Estimation Based on Efficient and Lightweight High-Resolution Network (EL-HRNet)
As an important direction in computer vision, human pose estimation has received extensive attention in recent years. A High-Resolution Network (HRNet) can achieve effective estimation results as a classical human pose estimation method. However, the complex structure of the model is not conducive t...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2024-01-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/24/2/396 |
| _version_ | 1797339423362252800 |
|---|---|
| author | Rui Li An Yan Shiqiang Yang Duo He Xin Zeng Hongyan Liu |
| author_facet | Rui Li An Yan Shiqiang Yang Duo He Xin Zeng Hongyan Liu |
| author_sort | Rui Li |
| collection | DOAJ |
| description | As an important direction in computer vision, human pose estimation has received extensive attention in recent years. A High-Resolution Network (HRNet) can achieve effective estimation results as a classical human pose estimation method. However, the complex structure of the model is not conducive to deployment under limited computer resources. Therefore, an improved Efficient and Lightweight HRNet (EL-HRNet) model is proposed. In detail, point-wise and grouped convolutions were used to construct a lightweight residual module, replacing the original 3 × 3 module to reduce the parameters. To compensate for the information loss caused by the network’s lightweight nature, the Convolutional Block Attention Module (CBAM) is introduced after the new lightweight residual module to construct the Lightweight Attention Basicblock (LA-Basicblock) module to achieve high-precision human pose estimation. To verify the effectiveness of the proposed EL-HRNet, experiments were carried out using the COCO2017 and MPII datasets. The experimental results show that the EL-HRNet model requires only 5 million parameters and 2.0 GFlops calculations and achieves an AP score of 67.1% on the COCO2017 validation set. In addition, PCKh@0.5mean is 87.7% on the MPII validation set, and EL-HRNet shows a good balance between model complexity and human pose estimation accuracy. |
| first_indexed | 2024-03-08T09:47:54Z |
| format | Article |
| id | doaj.art-92e487b7286e43698482ba59daf984bc |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-08T09:47:54Z |
| publishDate | 2024-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-92e487b7286e43698482ba59daf984bc2024-01-29T14:14:11ZengMDPI AGSensors1424-82202024-01-0124239610.3390/s24020396Human Pose Estimation Based on Efficient and Lightweight High-Resolution Network (EL-HRNet)Rui Li0An Yan1Shiqiang Yang2Duo He3Xin Zeng4Hongyan Liu5School of Mechanical and Precision Instrument Engineering, Xi’an University of Technology, Xi’an 710054, ChinaSchool of Mechanical and Precision Instrument Engineering, Xi’an University of Technology, Xi’an 710054, ChinaSchool of Mechanical and Precision Instrument Engineering, Xi’an University of Technology, Xi’an 710054, ChinaSchool of Mechanical and Precision Instrument Engineering, Xi’an University of Technology, Xi’an 710054, ChinaSchool of Mechanical and Precision Instrument Engineering, Xi’an University of Technology, Xi’an 710054, ChinaSchool of Mechanical and Precision Instrument Engineering, Xi’an University of Technology, Xi’an 710054, ChinaAs an important direction in computer vision, human pose estimation has received extensive attention in recent years. A High-Resolution Network (HRNet) can achieve effective estimation results as a classical human pose estimation method. However, the complex structure of the model is not conducive to deployment under limited computer resources. Therefore, an improved Efficient and Lightweight HRNet (EL-HRNet) model is proposed. In detail, point-wise and grouped convolutions were used to construct a lightweight residual module, replacing the original 3 × 3 module to reduce the parameters. To compensate for the information loss caused by the network’s lightweight nature, the Convolutional Block Attention Module (CBAM) is introduced after the new lightweight residual module to construct the Lightweight Attention Basicblock (LA-Basicblock) module to achieve high-precision human pose estimation. To verify the effectiveness of the proposed EL-HRNet, experiments were carried out using the COCO2017 and MPII datasets. The experimental results show that the EL-HRNet model requires only 5 million parameters and 2.0 GFlops calculations and achieves an AP score of 67.1% on the COCO2017 validation set. In addition, PCKh@0.5mean is 87.7% on the MPII validation set, and EL-HRNet shows a good balance between model complexity and human pose estimation accuracy.https://www.mdpi.com/1424-8220/24/2/396human pose estimationlightweight networkHRNetCBAM |
| spellingShingle | Rui Li An Yan Shiqiang Yang Duo He Xin Zeng Hongyan Liu Human Pose Estimation Based on Efficient and Lightweight High-Resolution Network (EL-HRNet) Sensors human pose estimation lightweight network HRNet CBAM |
| title | Human Pose Estimation Based on Efficient and Lightweight High-Resolution Network (EL-HRNet) |
| title_full | Human Pose Estimation Based on Efficient and Lightweight High-Resolution Network (EL-HRNet) |
| title_fullStr | Human Pose Estimation Based on Efficient and Lightweight High-Resolution Network (EL-HRNet) |
| title_full_unstemmed | Human Pose Estimation Based on Efficient and Lightweight High-Resolution Network (EL-HRNet) |
| title_short | Human Pose Estimation Based on Efficient and Lightweight High-Resolution Network (EL-HRNet) |
| title_sort | human pose estimation based on efficient and lightweight high resolution network el hrnet |
| topic | human pose estimation lightweight network HRNet CBAM |
| url | https://www.mdpi.com/1424-8220/24/2/396 |
| work_keys_str_mv | AT ruili humanposeestimationbasedonefficientandlightweighthighresolutionnetworkelhrnet AT anyan humanposeestimationbasedonefficientandlightweighthighresolutionnetworkelhrnet AT shiqiangyang humanposeestimationbasedonefficientandlightweighthighresolutionnetworkelhrnet AT duohe humanposeestimationbasedonefficientandlightweighthighresolutionnetworkelhrnet AT xinzeng humanposeestimationbasedonefficientandlightweighthighresolutionnetworkelhrnet AT hongyanliu humanposeestimationbasedonefficientandlightweighthighresolutionnetworkelhrnet |