How are the parameters of a convolutional neural network with a fully connected layer determined?

Convolutional neural network parameter determination with fully connected layers: Convolutional neural networks are different from traditional face detection methods in that they work directly on the input samples, which are used to train the network and ultimately achieve the detection task.

It is a non-parametric face detection method, which can eliminate a series of complex processes such as modeling, parameter estimation, and parameter testing, reconstruction of the model, etc. in traditional methods. In this paper, we target faces of any size, position, pose, orientation, skin color, facial expression and lighting conditions in the image.

Input layer

The input layer of a convolutional neural network can handle multidimensional data, commonly, the input layer of a one-dimensional convolutional neural network receives one- or two-dimensional arrays, where the one-dimensional arrays are usually temporal or spectral samples; the two-dimensional arrays may contain multiple channels; the input layer of a two-dimensional convolutional neural network receives two- or three-dimensional arrays; and the input layer of a three-dimensional convolutional neural network receives four-dimensional arrays.

Because convolutional neural networks are widely used in the field of computer vision, many studies have introduced their structure with the pre-assumption of three-dimensional input data, i.e., two-dimensional pixel points on a plane and RGB channels.