A Deep Convolutional Neural Network Model for Intelligent Discrimination between Coal and Rocks in Coal Mining Face

Table 1

Convolutional neural network structure.

Layer types	Output size	Patch size	Learning parameters
Input	(256 × 256 × 3)	—	Initial learning rate = 0.01
Conv1	(256 × 256 × 64)	(64, 3, 3)	Learning rate decay in each of 20 iterations = 50%
Conv2	(256 × 256 × 64)	(64, 3, 3)	Maximum number of iterations = 250
MaxPool1	(128 × 128 × 64)	(2, 2)
Conv3	(128 × 128 × 128)	(128, 3, 3)
Conv4	(128 × 128 × 128)	(128, 3, 3)
MaxPool2	(64 × 64 × 128)	(2, 2)
Inception1	(64 × 64 × 256)	—
MaxPool3	(32 × 32 × 256)	(2, 2)
Inception2	(32 × 32 × 256)	(512, 3, 3)
MaxPool4	(16 × 16 × 256)	(2, 2)
Inception3	(16 × 16 × 256)	(512, 3, 3)
MaxPool5	(8 × 8 × 256)	(2, 2)
Conv5	(8 × 8 × 512)	(512, 1, 1)
Flatten	(512 × 1 × 1)	—
Dense	(256 × 1 × 1)	—
Output	3	—