MULTI-SCALE CONTEXT AGGREGATION BY DILATED CONVOLUTIONS

采用dilated convolution代替pooling层,关于dilated conviolution的原理,可见: http://blog.csdn.net/u011961856/article/details/77141761 the front end module: 输入为三通道padded彩色图像,输出为输出feature map为​,output channels为21.采用VG

>>阅读原文<<