Figure 1.

The architecture of our newly developed network, called BibNet. Here, the blue boxes represent convolutional blocks with each block comprising a dropout layer, a convolutional layer, an activation layer and a batch normalization layer. The black boxes represent residual blocks; each comprising two convolutional blocks and a skip connection. The number inside the box represent the number of convolutional or residual blocks which are appended. The arrows indicate the connections in between the convolutional and residual blocks: the black arrows are forwarding the output of the previous block to the input of the next block, whereas the gray arrows symbolize a strided convolution, which is performed to decrease the resolution of the output of the previous layer. Additionally, the red arrows indicate transposed convolutions, which are used to double the resolution in each dimension.