Figure 1 –

General schema of a convolutional neural networks. Convolutional and max pooling layers can be stacked for deeper networks. The input image is convolved with a sliding window (yellow square), resulting in a set of feature maps that are processed by the max-pooling layers. The output of the final max-pooling is then processed by a set of dense layers responsible to interpret the feature maps and provide the estimated value of elasticity or viscosity as output.