Skip to main content
. 2018 May 21;8:7911. doi: 10.1038/s41598-018-26350-3

Figure 5.

Figure 5

A visualisation of the proposed Neural Network architecture, showing, in order, the following layers: MDLSTM, convolution, MDLSTM, and fully connected. In each MDLSTM layer, the four MDLSTM blocks, which process the input in four different directions, are depicted as four arrows highlighting the direction each processes pixels in.