[Retracted] Hybrid Depth-Separable Residual Networks for Hyperspectral Image Classification

Table 8

The comparison experiments in the IN dataset.

No. of classes	Train/test	SVM-RBF	2D CNN	3D CNN	SSRN	HybridSN	HDSRN
1	9/37	61.5	85.88	94.63	97.82	97.98	99.78
2	286/1142	78.68	91.31	93.9	99.17	98.37	99.59
3	166/664	73.41	91.07	94.85	99.53	99.48	99.82
4	47/190	71.58	80.38	93.48	97.79	97.38	99.82
5	97/386	80.38	91.89	93.56	99.24	99.23	99.38
6	146/584	92.27	99.01	94.2	99.51	99.14	99.89
7	6/22	79.52	82.59	89.73	98.7	99	99.32
8	96/382	87.38	100	96.01	99.85	100	100
9	4/16	85.87	66.55	95	98.5	99.01	100
10	194/778	77.58	86.38	94.55	98.74	98.76	99.61
11	491/1964	83.75	90.47	93.87	99.3	99.64	99.89
12	119/474	83.21	82.89	91.52	98.43	99.12	98.55
13	41/164	84.64	99.06	93.89	100	100	99.86
14	253/1012	98.01	97.86	91.77	99.31	100	100
15	77/309	94.3	90.52	95.03	99.2	99.35	99.89
16	19/74	61.43	98.94	93.57	97.82	97.26	98.57
OA (%)		82.83	90.89	94.07	99.19	99.26	99.72
AA (%)		80.84	89.68	93.72	98.93	98.98	99.62
Kappa × 100		82.23	88.56	93.87	99.07	99.09	99.70