Convolution Neural Network Based on Two-Dimensional Spectrum for Hyperspectral Image Classification

Table 6

Performance comparison of dropout, LRN, and BN.

Dataset	Operating layer	Unit time of batch/s	Error	Kappa
Indian pines	Dropout	0.940	0.0860	0.905
	LRN	0.898	0.0629	0.930
	B N	0.742	0.0174	0.978
Salinas	Dropout	0.671	0.0392	0.945
	LRN	0.360	0.0323	0.939
	B N	0.342	0.0272	0.972
KSC	Dropout	0.115	0.0893	0.892
	LRN	0.114	0.0592	0.922
	B N	0.099	0.0378	0.956
Botswana	Dropout	0.362	0.1709	0.829
	LRN	0.323	0.0970	0.886
	B N	0.283	0.0207	0.963

Owing to the number of samples between the datasets and the specific parameters of the training, this entry only applies to the comparison within the same dataset.