Towards Fine Whole-Slide Skeletal Muscle Image Segmentation through Deep Hierarchically Connected Networks

Figure 3

The detailed network configuration. The convolutional, max-pooling, deconvolutional, and concat layers are denoted by conv, pool, deconv, and concat, respectively. Each convolutional layer of the encoder is followed by a ReLU layer which is hidden in the tables. There are 5 decoders connected inside the architecture of the encoder. The (black solid and gray dotted) arrows point to the layer where the output of the corresponding layer goes. The last column of each table shows the feature map size (height ×  width ×  dimension) of each layer. In the tables of decoders, “†” indicates that a crop layer is connected after that to force the output size to be the same as the input image size (i.e., in the table).