Two-dimensional multi-human-body posture estimation method based on deep residual neural network

The invention discloses a two-dimensional multi-human-body posture estimation method based on a deep residual neural network, and the method comprises the steps: 1), obtaining a basic human body posture estimation training data set, 2), transmitting an obtained image to the pre-trained deep residual neural network, and generating a corresponding feature vector; 3) performing multiple convolution operations in two directions on the feature vector to generate feature mapping of two branches, and generating an articulation point confidence field and a part affinity field; 4) calculating an articulation point confidence field and a part affinity field of the real picture, combining the articulation point confidence field and the part affinity field into feature mapping of the real picture, andperforming network training on a mean square error between the feature mapping in the step 3) and the feature mapping of the real picture; repeating the steps 1)-3) to generate a joint point confidence field and a part affin