Table 8 Comparisons on model time of different methods on IP, KSC, UP, and HU datasets.
From: Adaptive pixel attention network for hyperspectral image classification
Machine learning | Deep learning | ||||||||
|---|---|---|---|---|---|---|---|---|---|
Data set | Metric | Context33 | RSSAN23 | SSTN31 | SSSAN40 | SSAtt24 | A2S2K25 | CVSSN19 | Proposed |
IP | train(s) | 135.32 | 9.89 | 25.28 | 46.49 | 11.34 | 223.35 | 22.96 | 20.73 |
test(ms) | 55.39 | 3.96 | 13.03 | 22.71 | 4.54 | 85.07 | 7.23 | 10.10 | |
KSC | train(s) | 76.71 | 5.34 | 13.81 | 26.09 | 5.94 | 112.25 | 12.26 | 15.73 |
test(ms) | 108.91 | 5.73 | 16.02 | 29.63 | 6.78 | 133.89 | 16.38 | 22.75 | |
UP | train(s) | 378.39 | 25.39 | 60.42 | 126.3 | 29.73 | 342.16 | 60.03 | 76.05 |
test(ms) | 28.82 | 1.96 | 6.14 | 16.23 | 3.18 | 25.44 | 3.60 | 5.39 | |
HU | train(s) | 123.7 | 8.76 | 21.99 | 51.99 | 10.25 | 234.71 | 19.57 | 22.39 |
test(ms) | 68.63 | 4.12 | 12.07 | 26.90 | 5.58 | 80.99 | 8.02 | 10.52 | |
