Paragraph Generation on Stanford Image Paragraph Dataset 1.0 (test)
19.22METEORHuman
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Human2016.11 | 19.22 | 28.55 | 42.88 | 25.68 | 15.55 | 9.66 | |
| CAE-LSTMBackbone=VGG16, Region Proposals=50, self-critical training=true2019.08 | 18.82 | 25.15 | — | — | — | 9.67 | |
| CapG-RevGBackbone=VGG16, Region Proposals=502019.08 | 18.62 | 20.93 | — | — | — | 9.43 | |
| RTT-GANBackbone=VGG16, Region Proposals=502019.08 | 18.39 | 20.36 | — | — | — | 9.21 | |
| LSTM-ATTBackbone=VGG16, Region Proposals=50, self-critical training=false2019.08 | 17.72 | 20.17 | — | — | — | 8.72 | |
| Regions-Hierarchical2016.11 | 15.95 | 13.52 | 41.9 | 24.11 | 14.23 | 8.69 | |
| Regions-HierarchicalBackbone=VGG16, Region Proposals=502019.08 | 15.95 | 13.52 | — | — | — | 8.69 | |
| Templatetype=heuristic/template-based2016.11 | 14.31 | 12.15 | 37.47 | 21.02 | 12.3 | 7.38 | |
| Regions-Flat-Pretrainedinitialization=pre-trained2016.11 | 14.23 | 12.13 | 38.32 | 22.9 | 14.17 | 8.97 | |
| Regions-Flat-Scratchinitialization=scratch2016.11 | 13.54 | 11.14 | 37.3 | 21.7 | 13.07 | 8.07 | |
| Image-Flat2016.11 | 12.82 | 11.06 | 34.04 | 19.95 | 12.2 | 7.71 | |
| Image-FlatBackbone=VGG16, Region Proposals=502019.08 | 12.82 | 11.06 | — | — | — | 7.71 | |
| DenseCap-Concat2016.11 | 12.66 | 12.51 | 33.18 | 16.92 | 8.54 | 4.54 | |
| Sentence-Concat2016.11 | 12.05 | 6.82 | 31.11 | 15.1 | 7.56 | 3.98 |