Table 5 Effect of the different labelling methods on the evaluation criteria values.
From: Remote sensing image description based on word embedding and end-to-end deep learning
Percentage—decoding layer | Spectrum | Texture | Fusion | Word vector | Original features |
|---|---|---|---|---|---|
UAV | |||||
P | 0.8631 | 0.8699 | 0.8837 | 0.9234 | 0.8960 |
R | 0.8372 | 0.8402 | 0.8563 | 0.8911 | 0.8692 |
F | 0.8499 | 0.8548 | 0.8698 | 0.9069 | 0.8824 |
Quick Bird | |||||
P | 0.8743 | 0.8755 | 0.8905 | 0.9057 | 0.8839 |
R | 0.8411 | 0.8436 | 0.8654 | 0.8741 | 0.8527 |
F | 0.8574 | 0.8593 | 0.8778 | 0.8896 | 0.8680 |
