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

Table 6

Classification results of different methods for the University of Pavia dataset.

Class	1D-CNN	M3D-DCNN	SC-FR	pResNet	DBDA	Proposed
1	74.36	95.32	99.79	99.51	99.80	99.60
2	71.64	97.20	99.58	99.69	99.73	99.58
3	82.51	94.84	99.62	99.82	99.69	99.72
4	95.01	98.47	97.14	99.16	96.54	99.74
5	99.71	99.95	99.44	99.99	99.50	100.00
6	83.08	98.51	99.52	99.98	99.97	99.95
7	93.72	97.69	100.00	99.98	99.94	100.00
8	80.81	95.85	99.06	99.79	99.41	99.85
9	99.82	99.71	99.14	99.89	99.20	99.96
OA (%)	77.85 ± 1.50	97.04 ± 1.33	99.40 ± 0.28	99.69 ± 0.11	99.52 ± 0.17	99.69 ± 0.15
Variance (OA)	2.2536	1.7582	0.0793	0.0132	0.0281	0.0226
AA (%)	86.74 ± 0.37	97.50 ± 0.87	99.26 ± 0.19	99.76 ± 0.07	99.31 ± 0.21	99.82 ± 0.06
Variance (AA)	0.1337	0.7536	0.0372	0.0053	0.0452	0.0037
KAPPA × 100	71.40 ± 1.73	95.97 ± 1.79	99.18 ± 0.39	99.58 ± 0.16	99.34 ± 0.23	99.58 ± 0.21