Abstract 1、提出a novel structure-aware convolutional network,能够在深度网络的训练过程中隐式地考虑先验知识(人体关节的几何约束关系); 2、Generator预测的pose和occlusion heatmaps输入到discriminator预测与真实姿态的相似度; 3、Conditional Generative Adversarial Networks(GANs)训练策略 目标:If the pose generator(G) generates results that the discriminator fails to dist