goolenet论文4篇 [v1]GoingDeeperwithConvolutions,6.67%testerror,http://arxiv.org/abs/1409.4842[v2]BatchNormalization:AcceleratingDeepNetworkTrainingbyReducingInternalC