Music genres are conventional categories that identify some pieces of music as belonging to a shared tradition or set of conventions. In this paper, we proposed an approach to improve music genre classification with convolutional neural networks (CNN). Using mel-scale spectrogram as the input, we used duplicate convolutional layers whose output will be applied to different pooling layers to provide more statistical information for classification. Also, we made some modifications on residual learning by taking more outputs from convolutional layers. By comparing two different network topologies, our experimental results on the GTZAN dataset show that the proposed method can effectively improve the classification accuracy.