Fig. 1
From: RODAN: a fully convolutional architecture for basecalling nanopore RNA sequencing data
The RODAN architecture. The normalized signal is passed through a succession of convolutional blocks which gradually incorporate surrounding information. Each block is composed of several processing steps (convolution, activation, batch normalization etc.), which are standard building blocks in the construction of deep neural networks. The final output is passed through a fully connected layer to produce the decoded sequence of nucleotides