Maximizing the Accuracy of Continuous Quantification Measures Using Discrete PackTest Products with Deep Learning and Pseudocolor Imaging

Figure 2

Deep learning architecture used in this study. Affine is a networking structure. Tanh is a process of hyperbolic tangent-conversion of the input. Sigmoid converts a value generated by the upstream processes. Sigmoid results in a sigmoid-converted value between 0 and 1. Squared Error is a process for minimizing errors in the estimation of the value.