Segment-level Video Captioning on ViTT Cooking (test)
41.61BLEU-1Human
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| HumanInput=Video + ASR2020.11 | 41.61 | 32.5 | 41.59 | 1.21 | |
| E2vidD6-MASSalign-BiDInput=Video + ASR, Pre-training strategy=Multimodal Pretraining2020.11 | 24.92 | 12.25 | 33.09 | 0.9 | |
| E2vidD6-MASSvid-BiDInput=Video + ASR, Pre-training strategy=Multimodal Pretraining2020.11 | 24.87 | 12.43 | 32.97 | 0.9 | |
| E2D6-MASS-BiDInput=ASR, Pre-training strategy=Text Pretraining2020.11 | 24.79 | 12.25 | 32.4 | 0.88 | |
| E2vidD6-MASSdrop-BiDInput=Video + ASR, Pre-training strategy=Multimodal Pretraining2020.11 | 24.48 | 12.22 | 33.1 | 0.89 | |
| E2vidD6-MASS-BiDInput=Video + ASR, Pre-training strategy=Text Pretraining2020.11 | 24.22 | 12.22 | 32.6 | 0.89 | |
| E2D6-BiDInput=ASR, Pre-training strategy=Ø Pretraining2020.11 | 20.77 | 10.08 | 28.63 | 0.72 | |
| E2vidD6-BiDInput=Video + ASR, Pre-training strategy=Ø Pretraining2020.11 | 20.45 | 9.88 | 28.88 | 0.69 | |
| Constant baseline (“intro”)Input=-2020.11 | 1.16 | 2.93 | 10.21 | 0.25 |