Table 1 Comparison scores of our method and other state-of-the-art methods on the NWPU-Captions dataset.
From: Feature refinement and rethinking attention for remote sensing image captioning
Methods | BLEU1 | BLEU2 | BLEU3 | BLEU4 | METEOR | ROUGE_L | CIDEr |
|---|---|---|---|---|---|---|---|
SAT | 0.7340 | 0.6120 | 0.5280 | 0.4690 | 0.3370 | 0.6010 | 1.1090 |
SM-Att | 0.7390 | 0.6160 | 0.5340 | 0.4690 | 0.3380 | 0.5950 | 1.1370 |
Struc-Att | 0.7442 | 0.6091 | 0.5193 | 0.4557 | 0.3087 | 0.6064 | 1.2270 |
MLCA-Net | 0.7450 | 0.6240 | 0.5410 | 0.4780 | 0.3370 | 0.6010 | 1.2640 |
MC-Net | 0.7410 | 0.6260 | 0.5440 | 0.4780 | 0.3470 | 0.6110 | 1.1590 |
VRTMM | 0.8116 | 0.7033 | 0.6213 | 0.5570 | 0.3660 | 0.6845 | 1.5885 |
GLCM | 0.5536 | 0.4228 | 0.3353 | 0.2720 | 0.2789 | 0.5042 | 1.2774 |
P-to-H | 0.7571 | 0.6291 | 0.5457 | 0.4828 | 0.3187 | 0.5858 | 1.2071 |
DiffNet | 0.7972 | 0.6635 | 0.5604 | 0.4793 | 0.3059 | 0.6182 | 1.2324 |
Ours | 0.8490 | 0.7620 | 0.6957 | 0.6441 | 0.4198 | 0.7468 | 1.8167 |
