Spectral-Spatial Attention Transformer with Dense Connection for Hyperspectral Image Classification

Table 8

Classification results of different methods for the Salinas dataset.

Class	1D-CNN	M3D-DCNN	SC-FR	pResNet	DBDA	Proposed
1	84.89	99.85	99.75	100.00	100.00	100.00
2	98.17	99.97	99.80	100.00	100.00	100.00
3	89.97	99.85	99.13	99.93	100.00	100.00
4	99.51	99.63	89.39	99.97	99.69	99.74
5	97.06	97.42	99.39	99.83	97.46	99.72
6	99.76	99.99	100.00	100.00	100.00	100.00
7	99.42	99.62	100.00	99.95	100.00	99.99
8	84.76	93.23	97.23	99.21	99.39	99.21
9	98.00	99.85	99.95	99.99	100.00	100.00
10	87.53	98.17	99.43	99.51	99.69	99.84
11	74.88	98.80	98.92	99.75	99.97	99.82
12	99.50	100.00	99.99	99.99	100.00	99.98
13	97.88	99.85	99.93	99.94	100.00	99.99
14	91.10	98.81	99.73	99.98	99.83	99.80
15	56.75	88.67	96.44	98.88	96.48	99.09
16	88.18	98.60	99.69	99.90	99.99	99.91
OA (%)	87.65 ± 0.48	96.67 ± 0.58	98.51 ± 0.71	99.63 ± 0.24	99.24 ± 0.75	99.67 ± 0.08
Variance (OA)	0.2272	0.3416	0.5099	0.0554	0.5635	0.0057
AA (%)	90.46 ± 0.65	98.27 ± 0.29	98.67 ± 1.79	99.80 ± 0.13	99.53 ± 0.34	99.82 ± 0.05
Variance (AA)	0.4164	0.0817	3.1987	0.0171	0.1147	0.0029
KAPPA × 100	86.21 ± 0.53	96.29 ± 0.65	98.34 ± 0.80	99.59 ± 0.26	99.16 ± 0.84	99.63 ± 0.08