Research Article
Context-Fused Guidance for Image Captioning Using Sequence-Level Training
Table 4
Performance comparison of the ablative models.
| | Model | Cross-entropy training | CIDEr optimization | | Metric | BLEU4 | CIDEr | SPICE | BLEU4 | CIDEr | SPICE |
| | CFGV | 36.1 | 112.8 | 20.3 | 37.7 | 123.9 | 21.0 | | CFGE | 36.1 | 112.9 | 20.5 | 37.8 | 124.6 | 21.1 | | CFGA | 36.3 | 113.0 | 20.6 | 38.1 | 124.6 | 21.4 | | CFG | 36.8 | 114.0 | 20.8 | 38.3 | 125.4 | 21.6 |
|
|
