如何从预测图像中获取标签/类别

Question

我已经使用 cnn 训练了一个分类器模型以及如何对图像进行分类，使用 flowfromdirectory 加载图像。训练后，我对看不见的数据进行了测试，然后我的预测值是这样的：-

[[[[0.771866   0.8567589  0.902191  ]
   [0.8756012  0.93743145 0.9572767 ]
   [0.8575856  0.92234766 0.95071596]
   ...
   [0.8285682  0.9203037  0.9460905 ]
   [0.85615414 0.9182381  0.9420978 ]
   [0.7868526  0.8462486  0.8882149 ]]

  [[0.815824   0.8986912  0.93497586]
   [0.88749367 0.94393086 0.96703947]
   [0.8745645  0.93561155 0.9639484 ]
   ...
   [0.85607237 0.93997157 0.96361405]
   [0.8750243  0.937056   0.9653714 ]
   [0.8367853  0.89637643 0.9370449 ]]

  [[0.695155   0.7554943  0.77888066]
   [0.8337961  0.8617642  0.8727031 ]
   [0.8334781  0.84559196 0.86962295]
   ...
   [0.01496893 0.03932272 0.01421419]
   [0.01424129 0.04037565 0.01449471]
   [0.02615304 0.05672317 0.0271827 ]]

  [[0.71646017 0.76629317 0.79345727]
   [0.8340155  0.86742646 0.87385976]
   [0.84214795 0.8550026  0.872459  ]
   ...
   [0.01343644 0.0353599  0.01344425]
   [0.01343626 0.03432167 0.01197887]
   [0.02437622 0.05071671 0.02522665]]

如何将其转换为 class_indiceis/labels？