A Better Variant of Self-Critical Sequence Training
Open paper
Benchmark
Model
Rank
Results
image-captioning-on-coco-captions
Transformer_NSC
#23
BLEU-4: 39.4
CIDER: 129.6
METEOR: 28.9
SPICE: 22.8
ROUGE-L: 58.7
…
Rank counts only results with a code link.