ABSTRACT
The human visual cortex is segmented into small clusters of neurons that are responsive to specific features or sectors in the visual field, these so-called receptive fields are then merged to constitute the entire mental image in the neural network. The human visual cortex is segmented into small clusters of neurons that are responsive to specific features or sectors in the visual field, these so-called receptive fields are then merged to constitute the entire mental image in the neural network. In a multilayer perceptron neural network, all the neurons in adjacent layers are fully connected, and therefore cannot directly pick out and focus on specific areas of the visual field for closer examination. A convolutional filter striding over an artificial neural network matrix layer can be seen as a sliding inner product that extracts the confluences between the filter and the matrix layer, detecting, augmenting, or dampening, and thereby extracting features.
