Recently,remarkableadvanceshavebeenachievedin3DhumanposeestimationfrommonocularimagesbecauseofthepowerfulDeepConvolutionalNeuralNetworks(DC-NNs).Despitetheirsuccessonlarge-scaledatasetscol-lectedintheconstrainedlabenvironment,itisdifficulttoobtainthe3Dposeannotationsforin-the-wildimage