Transform-based human body posture estimation method

The invention discloses a human body posture estimation method based on Transform. According to the method, global and local features are extracted through a CNN and Transform combined double-branch network. Afterwards, the human body key points are directly regressed by adopting a'human body center-limb-key point 'multi-stage regression method; compared with a traditional detection method, the method has the advantages that global features and local features can be considered at the same time, and the defects that CNN ignores the global features and Transform ignores the local features are overcome; meanwhile, by adopting a multi-stage regression method, a'limb 'intermediate point is increased, the regression difficulty of a key point far away from the center of the human body is reduced, the regression precision of the key point is improved, and the estimation accuracy of the human body posture is further improved. 本发明公开了一种基于Transformer的人体姿态估计方法。该方法通过CNN与Transformer结合的双分支网络，提取全局与局部特征。随后，采用"人体中心-肢干-关键点"多阶段回归