An Axial Compression Transformer for Efficient Human Pose Estimation
Source abstract
Abstract
Transformer has a wide range of applications in human posture estimation. It can model the global dependence relationship of images through the self-attention mechanism to obtain key human body information. However, Transformer consumes a lot of computation. An axial compression pose transformer (ACPose) method is proposed to reduce part of the computational cost of Transformer by the axial compression of the input matrix, while maintaining the global receptive field by feature fusion. A Local Enhancement Module is constructed to avoid the loss of too much feature information in the compression process. In the COCO dataset experiment, there was a significant reduction in computational cost compared to those of state-of-the-art transformer-based algorithms.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Gait Recognition and Analysis · Non-Invasive Vital Sign Monitoring · Human Pose and Action Recognition
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
School of Computer Science and Technology, Zhejiang Sci-Tech University, Hangzhou 310018, China
Funding
Zhejiang Sci-Tech University Research Startup Funding Project
References (34)
Show source-provided references
1. Luvizon, D.C., Picard, D., and Tabia, H. (2018, January 18–23). 2D/3D Pose Estimation and Action Recognition Using Multitask Deep Learning. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, UT, USA.10.1109/CVPR.2018.00539 ↗
2. Geng, Z., Wang, C., Wei, Y., Liu, Z., Li, H., and Hu, H. (2023, January 18–22). Human Pose as Compositional Tokens. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA.10.1109/CVPR52729.2023.00071 ↗
3. BCCLR: A Skeleton-Based Action Recognition with Graph Convolutional Network Combining Behavior Dependence and Context Clues
4. Xu, X., and Zhang, Y. (2024). High-resolution multi-scale feature fusion network for running posture estimation. Appl. Sci., 14.10.3390/app14073065 ↗
5. PCNet: A Human Pose Compensation Network Based on Incremental Learning for Sports Actions Estimation10.1007/s40747-024-01647-1 ↗
6. Multi-task deep learning for real-time 3D human pose estimation and action recognition
7. Human–machine interaction and implementation on the upper extremities of a humanoid robot10.1007/s42452-024-05734-3 ↗
8. Azhar, M.H., and Jalal, A. (2024, January 26–27). Human-Human Interaction Recognition Using Mask R-CNN and Multi-Class SVM. Proceedings of the 2024 3rd International Conference on Emerging Trends in Electrical, Control, and Telecommunication Engineering (ETECTE), Lahore, Pakistan.10.1109/ETECTE63967.2024.10823924 ↗
9. Hernández, Ó.G., Morell, V., Ramon, J.L., and Jara, C.A. (2021). Human pose detection for robotic-assisted and rehabilitation environments. Appl. Sci., 11.10.3390/app11094183 ↗
10. A deep learning-enabled visual-inertial fusion method for human pose estimation in occluded human-robot collaborative assembly scenarios10.1016/j.rcim.2024.102906 ↗
11. Human Arm Pose Estimation with a Shoulder-worn Force-Myography Device for Human-Robot Interaction10.1109/LRA.2025.3539545 ↗
12. Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., and Gelly, S. (2021, January 4). An image is worth 16×16 words: Transformers for image recognition at scale. Proceedings of the International Conference on Learning Representations ICLR 2021, Vienna, Austria.
13. Dong, C., and Du, G. (2024). An enhanced real-time human pose estimation method based on modified YOLOv8 framework. Sci. Rep., 14.10.1038/s41598-024-58146-z ↗
14. Xia, H., and Zhang, T. (2021). Self-attention network for human pose estimation. Appl. Sci., 11.10.3390/app11041826 ↗
15. Fang, H.S., Xie, S., Tai, Y.W., and Lu, C. (2017, January 22–29). RMPE: Regional Multi-Person Pose Estimation. Proceedings of the IEEE International Conference on Computer Vision (ICCV), Venice, Italy.10.1109/ICCV.2017.256 ↗
16. Xiao, B., Wu, H., and Wei, Y. (2018, January 8–14). Simple Baselines for Human Pose Estimation and Tracking. Proceedings of the European Conference on Computer Vision (ECCV), Munich, Germany.10.1007/978-3-030-01231-1_29 ↗
17. Liu, Z., Lin, W., Shi, Y., and Zhao, J. (2021, January 13–15). A Robustly Optimized BERT Pre-Training Approach with Post-Training. Proceedings of the China National Conference on Chinese Computational Linguistics (CNCL), Hohhot, China.10.1007/978-3-030-84186-7_31 ↗
18. Exploring the limits of transfer learning with a unified text-to-text transformer
19. HRPVT: High-Resolution Pyramid Vision Transformer for Medium and Small-Scale Human Pose Estimation10.1016/j.neucom.2024.129154 ↗
20. Vision-Based Body Pose Estimation of Excavator Using a Transformer-Based Deep-Learning Model10.1061/JCCEE5.CPENG-6079 ↗
21. PCDPose: Enhancing the lightweight 2D human pose estimation model with pose-enhancing attention and context broadcasting10.1007/s10044-025-01431-y ↗
22. Yang, S., Quan, Z., Nie, M., and Yang, W. (2021, January 10–17). Transpose: Keypoint Localization via Transformer. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), Montreal, QC, Canada.10.1109/ICCV48922.2021.01159 ↗
23. Li, Y., Zhang, S., Wang, Z., Yang, S., Yang, W., Xia, S.T., and Zhou, E. (2021, January 10–17). TokenPose: Learning Keypoint Tokens for Human Pose Estimation. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), Montreal, QC, Canada.10.1109/ICCV48922.2021.01112 ↗
24. SharpPose: Sparse High-Resolution Representation for Human Pose Estimation10.1609/aaai.v38i2.27826 ↗
25. Tu, H., Qiu, Z., Yang, K., Tan, X., and Zheng, X. (2025). HP-YOLO: A Lightweight Real-Time Human Pose Estimation Method. Appl. Sci., 15.10.3390/app15063025 ↗
26. Cai, M., Jeon, W.S., and Rhee, S.Y. (2025, January 16–17). LW-FastPose: A Lightweight Network for Human Pose Estimation Based on Improvements to FastPose. Proceedings of the 2025 International Conference on Electronics, Information, and Communication (ICEIC), Rome, Italy.10.1109/ICEIC64972.2025.10879683 ↗
27. Noh, W.J., Moon, K.R., and Lee, B.D. (2024). SMS-Net: Bridging the Gap Between High Accuracy and Low Computational Cost in Pose Estimation. Appl. Sci., 14.10.3390/app142210143 ↗
28. Bao, H., Dong, L., Piao, S., and Wei, F. (2021). Beit: Bert pre-training of image transformers. arXiv.
29. Wan, Q., Huang, Z., Lu, J., Yu, G., and Zhang, L. (2023, January 1–5). SeaFormer: Squeeze-Enhanced Axial Transformer for Mobile Semantic Segmentation. Proceedings of the Eleventh International Conference on Learning Representations (ICLR), Kigali, Rwanda.
30. Dilateformer: Multi-scale dilated transformer for visual recognition10.1109/TMM.2023.3243616 ↗
31. Mao, W., Ge, Y., Shen, C., Tian, Z., Wang, X., and Wang, Z. (2021). Tfpose: Direct human pose estimation with transformers. arXiv.10.1007/978-3-031-20068-7_5 ↗
32. Vitpose: Simple vision transformer baselines for human pose estimation
33. Hrformer: High-resolution vision transformer for dense predict
34. Sun, K., Xiao, B., Liu, D., and Wang, J. (2019, January 15–20). Deep High-Resolution Representation Learning for Human Pose Estimation. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA.10.1109/CVPR.2019.00584 ↗