Fig. 1. Temporal pose estimation network^[18]

Fig. 2. CNN structure of detection-following-regression^[19]

Fig. 3. Logarithmic polarization grouping

Fig. 4. Multi-context attention mechanism framework^[22]

Fig. 5. Framework of generative adversarial networks^[26]

Fig. 6. Comparison of the results with and without human structure prior during network training^[27]

Fig. 7. Adversarial PoseNet model structure^[27]

Fig. 8. PRM structure^[28]

Fig. 9. Multi-scale structure-aware network^[11]

Fig. 10. Specific process of joint human body association^[44]

Fig. 11. Schematic of dilated convolution

Fig. 12. CPN overall structure^[46]

Fig. 13. Comparison of experimental results of various factors affecting the performance of multi-person pose estimation^[46]

Fig. 14. Human detection problem^[47]

Fig. 15. Pose estimation process of SPLP network^[38]

Fig. 16. Two-branch multi-stage CNN architecture^[48]

Fig. 17. Schematic of PAFs method^[48]

Fig. 18. Comparison results of single person pose estimation experiments (LSP)

Fig. 19. Comparison results of single person pose estimation experiments (MPII)

Fig. 20. Comparison results of multi-person pose estimation experiments (MPII)