Remote Sensing Image Water Body Recognition Algorithm Based on Deep Convolution Generating Network and Combined Features

Table 4

Initialization settings of water body identification model parameters.

Number of input layer nodes (feature dimension)	64
Number of output layer nodes	2 (water and nonwater)
Generate model ()	Learning rate	0.005
	Momentum	0.9
Discriminant model ()	Learning rate	0.005
	Momentum	0.9
Number of iterations	1000
Activation function	tanh
Classification function	SoftMax
Convolution kernel size	5
Pool layer filter size	2
Number of neurons in full connection layer	1024