A hyperspectral image classification algorithm based on atrous convolution

EURASIP Journal on Wireless Communications and Networking

Table 4 Classification results obtained by different approaches on Pavia University dataset

Classifier methods	OA (%)	Kappa	Parameters (M)	Memory (M)	Runtime (s)
SVM [7]	76.192	0.634			565
1D CNN [28]	84.619	0.792	0.23	0.25	255
1D CNN [29]	92.778	0.904	0.02	0.11	8519
2D CNN [30]	69.635	0.643	8.75	12.07	18148
3D CNN [31]	94.904	0.934	0.34	1.32	1062
3D CNN [32]	85.320	0.818	4.08	20.69	21313
DL CNN [33]	95.430	0.941	0.29	1.17	558
DSS CNN [34]	93.432	0.913	64.91	64.98	345
RNN [35]	94.460	0.927	0.30	0.33	219
NG-APC	97.966	0.973	0.05	0.25	433