Fig. 2
From: Nuclei instance segmentation from histopathology images using Bayesian dropout based deep learning
The convolution block of the BayesNuSeg model. The encoder unit contains five convolution blocks. The two independent decoder units contain the same number of convolution blocks, where a transpose convolution operation replaces the convolution operation. The parameter k represents the kernel size, s and p denote stride and padding, respectively. Finally, the dropout layer with a probability of 0.5 is activated during the inference phase to estimate the model uncertainty