Table 2.

Detailed configuration of the three datasets (Pines, University of Pavia (PaviaU) and Salinas) used in this paper. For each class, 200 samples are randomly selected as the training set and the rest as the testing set. For Pines, we use only the top 9 classes with the largest number of samples.

No.	Pines			PaviaU			Salinas
	Class Name	Train	Test	Class Name	Train	Test	Class Name	Train	Test
1	Corn-notill	200	1228	Asphalt	200	6431	Brocoli_1	200	1809
2	Corn-mintill	200	630	Meadows	200	18,449	Brocoli_2	200	3526
3	Grass-pasture	200	283	Gravel	200	1899	Fallow	200	1776
4	Grass-trees	200	530	Trees	200	2864	Fallow_plow	200	1194
5	Hay-win.	200	278	Sheets	200	1145	Fallow_smooth	200	2478
6	Soy.-notill	200	772	Bare Soil	200	4829	Stubble	200	3759
7	Soy.-mintill	200	2255	Bitumen	200	1130	Celery	200	3379
8	Soy.-clean	200	393	Bricks	200	3482	Grapes	200	11,071
9	Woods	200	1065	Shadows	200	747	Soil_vinyard	200	6003
10							Corn_weeds	200	3078
11							Lettuce_4wk	200	868
12							Lettuce_5wk	200	1727
13							Lettuce_6wk	200	716
14							Lettuce_7wk	200	870
15							Vinyard_un.	200	7068
16							Vinyard_ve.	200	1607
Sum		1800	7434		1800	40,976		3200	50,929