GRIT (No VL pretraining - base)

Reported on 6 benchmarks across 1 task · 1 paper · 2 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Natural Language Processing6 results

Image CaptioningonCOCO Captions
BLEU-1· 2022-07-20
84.2
SOTA
GRIT: Faster and Better Image captioning Transformer Using Dual Visual Features arXiv:2207.09666
Image CaptioningonCOCO Captions
ROUGE-L· 2022-07-20
60.7
best: 61.1 (ExpansionNet v2 (No VL pretraining))
SOTA
GRIT: Faster and Better Image captioning Transformer Using Dual Visual Features arXiv:2207.09666
Image CaptioningonCOCO Captions
BLEU-4· 2022-07-20
42.4
best: 46.5 (mPLUG)
GRIT: Faster and Better Image captioning Transformer Using Dual Visual Features arXiv:2207.09666
Image CaptioningonCOCO Captions
CIDER· 2022-07-20
144.2
best: 155.1 (mPLUG)
GRIT: Faster and Better Image captioning Transformer Using Dual Visual Features arXiv:2207.09666
Image CaptioningonCOCO Captions
METEOR· 2022-07-20
30.6
best: 33.9 (CoCa)
GRIT: Faster and Better Image captioning Transformer Using Dual Visual Features arXiv:2207.09666
Image CaptioningonCOCO Captions
SPICE· 2022-07-20
24.3
best: 27 (VAST)
GRIT: Faster and Better Image captioning Transformer Using Dual Visual Features arXiv:2207.09666