Skip to content

Latest commit

 

History

History
37 lines (30 loc) · 1.92 KB

README.md

File metadata and controls

37 lines (30 loc) · 1.92 KB

SSRN

This is a keras reprodction of TRGS paper:Spectral–Spatial Residual Network for Hyperspectral Image Classification: A 3-D Deep Learning Framework

Test result on Indian Pines Dataset using the author's best parameter configurations:

Test loss: 0.14532101154327393
Test acc: 95.90243697166443%

Classification result:

                          precision    recall  f1-score   support

                 Alfalfa       1.00      0.50      0.67        32
             Corn-notill       0.96      0.92      0.94      1000
            Corn-mintill       0.96      0.98      0.97       581
                    Corn       0.95      0.99      0.97       166
           Grass-pasture       0.96      0.96      0.96       338
             Grass-trees       1.00      0.98      0.99       511
     Grass-pasture-mowed       1.00      0.47      0.64        19
           Hay-windrowed       0.92      1.00      0.96       334
                    Oats       1.00      0.93      0.96        14
          Soybean-notill       0.93      0.94      0.94       681
         Soybean-mintill       0.98      0.97      0.97      1719
           Soybean-clean       0.92      0.96      0.94       416
                   Wheat       1.00      0.99      0.99       143
                   Woods       0.97      0.98      0.97       886
Building-Gras-Tree-Drive       0.91      0.91      0.91       270
      Stone-Steel-Towers       0.97      1.00      0.98        65
      
      
                accuracy                           0.96      7175
               macro avg       0.96      0.90      0.92      7175
            weighted avg       0.96      0.96      0.96      7175

Confusion matrix: CM

Predict map: PM