Fig 2. Creation of the 3-D convolutional neural network.

a) Agreement matrix between six individuals (members of the authors’ research group), evaluated on a single dataset of images of Vibrio bacteria, and between those humans and the convolutional neural network. b) Accuracy vs number of kernels per layer using cross validation across the various imaging datasets, where the x-axis denotes the number of kernels in the first convolutional layer. The second convolutional layer for each plotted point has twice as many kernels as the first.