Classification Modeling Method for Near-Infrared Spectroscopy of Tobacco Based on Multimodal Convolution Neural Networks

Table 2

Simple 1-D CNN structures of different depths.

Model	Hidden layer structure	Feature map structure	Depth
1D_CNN_5	Stage1: C1 + S1	M1	5
1D_CNN_7	Stage1: C1 + S1; Stage2: C2 + S2	(M1, M2)	7
1D_CNN_9	Stage1: C1 + S1; Stage2: C2 + S2; Stage3: C3 + S3	(M1, M2, M3)	9
1D_CNN_11	Stage1: C1 + S1; Stage2: C2 + S2; Stage3: C3 + S3; Stage4: C4 + S4	(M1, M2, M3, M4)	11
1D_CNN_13	Stage1: C1 + S1; Stage2: C2 + S2; Stage3: C3 + S3; Stage4: C4 + S4; Stage5: C5 + S5	(M1, M2, M3, M4, M5)	13