Table 4.

The overall accuracy of different models in University of Pavia.

Model\Amount of Training Samples	5%	2%	1.6%	1.2%	0.8%	0.4%
2D-CNN	96.59 ± 0.21	94.50 ± 0.40	93.55 ± 0.22	91.82 ± 0.56	89.98 ± 0.38	85.27 ± 0.90
M3D-DCNN	92.80 ± 0.95	89.27 ± 1.35	88.28 ± 1.47	87.19 ± 1.71	82.75 ± 2.84	76.53 ± 3.94
HybridSN	99.45 ± 0.09	97.86 ± 0.56	96.87 ± 0.31	95.86 ± 0.93	93.30 ± 1.41	85.95 ± 1.58
Model A	99.37 ± 0.12	97.99 ± 0.53	97.83 ± 0.64	95.01 ± 1.44	93.98 ± 1.70	90.85 ± 1.51
Model B	99.03 ± 0.10	96.89 ± 0.59	96.45 ± 1.02	88.54 ± 4.45	91.05 ± 2.33	81.06 ± 3.21
R-HybridSN	99.47 ± 0.14	98.47 ± 0.27	98.30 ± 0.21	96.40 ± 1.66	95.64 ± 0.52	91.60 ± 1.12