Caption-level correlation with human judgment on Composite (test)
0.6Kendall's TauRefCLIP-S
Evaluation Results
| Method | Links | |
|---|---|---|
| RefCLIP-Sevaluation_setting=easier setting, number_of_references=4-52021.04 | 0.6 | |
| RefCLIP-Sevaluation_setting=harder setting, number_of_references=4-52021.04 | 0.554 | |
| CLIP-S (no refs)reference_usage=none2021.04 | 0.538 | |
| ViLBERTScore-Fevaluation_setting=harder setting, number_of_references=4-52021.04 | 0.524 | |
| TIGErevaluation_setting=harder setting, number_of_references=4-52021.04 | 0.454 | |
| BERT-S++evaluation_setting=harder setting, reported_in_prior_work=true, number_of_references=4-52021.04 | 0.449 | |
| SPICEEvaluation Level=Caption-level2021.06 | 0.445 | |
| SPARCSEvaluation Level=Caption-level2021.06 | 0.431 | |
| CIDErEvaluation Level=Caption-level2021.06 | 0.407 | |
| METEOREvaluation Level=Caption-level2021.06 | 0.404 | |
| SPICEevaluation_setting=harder setting, number_of_references=4-52021.04 | 0.403 | |
| Bleu-2Evaluation Level=Caption-level2021.06 | 0.394 | |
| Rouge-LEvaluation Level=Caption-level2021.06 | 0.393 | |
| METEORevaluation_setting=harder setting, number_of_references=4-52021.04 | 0.389 | |
| BERTScoreEvaluation Level=Caption-level2021.06 | 0.388 | |
| Bleu-1Evaluation Level=Caption-level2021.06 | 0.386 | |
| CIDErevaluation_setting=harder setting, number_of_references=4-52021.04 | 0.377 | |
| ROUGE-Levaluation_setting=harder setting, number_of_references=4-52021.04 | 0.324 | |
| BLEU-1evaluation_setting=harder setting, number_of_references=4-52021.04 | 0.313 | |
| BLEU-4evaluation_setting=harder setting, number_of_references=4-52021.04 | 0.306 | |
| BERT-S (RoBERTa-F)evaluation_setting=harder setting, number_of_references=4-52021.04 | 0.301 |