Caption Retrieval on MS COCO Karpathy 1k (test)
84R@1Discrete-continuous PG
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Discrete-continuous PGNetwork=Faster R-CNN2021.04 | 84 | 95.8 | 97.8 | — | — | |
| VSRNBackbone=Faster R-CNN, ResNet2019.09 | 76.2 | 94.8 | 98.2 | — | — | |
| VSRNNetwork=Faster R-CNN2021.04 | 76.2 | 94.8 | 98.2 | — | — | |
| SCANBackbone=Faster R-CNN, ResNet2019.09 | 72.7 | 94.8 | 98.4 | — | — | |
| SCANNetwork=Faster R-CNN2021.04 | 72.7 | 94.8 | 98.4 | — | — | |
| CAMP2019.09 | 72.3 | 94.8 | 98.3 | — | — | |
| SCO2019.09 | 69.9 | 92.9 | 97.5 | — | — | |
| SCOBackbone=ResNet2019.09 | 69.9 | 92.9 | 97.5 | — | — | |
| SCONetwork=ResNet2021.04 | 69.9 | 92.9 | 97.5 | — | — | |
| SCANMode=i-t2019.09 | 69.2 | 93.2 | 97.5 | — | — | |
| GXN2019.09 | 68.5 | — | 97.9 | — | — | |
| GXNBackbone=ResNet2019.09 | 68.5 | — | 97.9 | — | — | |
| GXNNetwork=ResNet2021.04 | 68.5 | — | 97.9 | — | — | |
| SCANMode=t-i2019.09 | 67.5 | 92.9 | 97.6 | — | — | |
| DPC2019.09 | 65.6 | 89.8 | 95.5 | — | — | |
| VSE++2019.09 | 64.6 | — | 95.7 | — | — | |
| VSE++Backbone=ResNet2019.09 | 64.6 | 89.1 | 95.7 | — | — | |
| VSENetwork=ResNet2021.04 | 64.6 | 89.1 | 95.7 | — | — | |
| CAMP2019.09 | 58.5 | 87.9 | 95 | — | — | |
| SCO2019.09 | 56.7 | 87.5 | 94.8 | — | — | |
| GXN2019.09 | 56.6 | — | 94.5 | — | — | |
| RRFBackbone=ResNet2019.09 | 56.4 | 85.3 | 91.5 | — | — | |
| RRFNetwork=ResNet2021.04 | 56.4 | 85.3 | 91.5 | — | — | |
| CMPM2019.09 | 56.1 | 86.3 | 92.9 | — | — | |
| 2WayNetBackbone=VGG2019.09 | 55.8 | — | 75.2 | — | — | |
| 2wayNNetwork=VGG2021.04 | 55.8 | 75.2 | — | — | — | |
| SCANMode=i-t2019.09 | 54.4 | 86 | 93.6 | — | — | |
| SM-lstmBackbone=VGG2019.09 | 53.2 | 83.1 | 91.5 | — | — | |
| SMlstmNetwork=VGG2021.04 | 53.2 | 83.1 | 91.5 | — | — | |
| SCANMode=t-i2019.09 | 53 | 85.4 | 92.9 | — | — | |
| VSE++2019.09 | 52 | — | 92 | — | — | |
| VQA-EBackbone=VGG2019.09 | 50.5 | 80.1 | 89.7 | — | — | |
| VQANetwork=VGG2021.04 | 50.5 | 80.1 | 89.7 | — | — | |
| DPC2019.09 | 47.1 | 79.9 | 90 | — | — | |
| order-embeddings2015.11 | 46.7 | — | 88.9 | 2 | 5.7 | |
| Order-embeddingsSupervision type=Direct supervision2017.05 | 46.7 | — | 88.9 | 2 | — | |
| Order2019.09 | 46.7 | — | 88.9 | — | — | |
| OEMBackbone=VGG2019.09 | 46.7 | — | 88.9 | — | — | |
| order-embeddingsVariant=symmetric2015.11 | 45.4 | — | 88.7 | 2 | 5.8 | |
| CMPM2019.09 | 44.6 | 78.8 | 89 | — | — | |
| HM-lstmBackbone=R-CNN, AlexNet2019.09 | 43.9 | — | 87.8 | — | — | |
| HMlstmNetwork=AlexNet2021.04 | 43.9 | — | 87.8 | — | — | |
| MNLM2015.11 | 43.4 | — | 85.8 | 2 | — | |
| m-CNN ENSEnsemble=true2015.11 | 42.8 | — | 84.1 | 2 | — | |
| m-CNNENSSupervision type=Direct supervision2017.05 | 42.8 | — | 84.1 | 2 | — | |
| BiLSTM-Max (on AllNLI)Supervision type=Pre-trained, Backbone=ResNet101, Training samples count=113k, Pre-training dataset=AllNLI2017.05 | 42.6 | 75.3 | 87.3 | 2 | — | |
| BiLSTM-Max (on SNLI)Supervision type=Pre-trained, Backbone=ResNet101, Training samples count=113k, Pre-training dataset=SNLI2017.05 | 42.4 | 76.1 | 87 | 2 | — | |
| order-embeddingsFeatures=1-crop VGG2015.11 | 41.4 | — | 84.2 | 2 | 8.7 | |
| m-RNN2015.11 | 41 | — | 83.5 | 2 | — | |
| FV2015.11 | 39.4 | — | 80.9 | 2 | 10.4 | |
| FVBackbone=VGG2019.09 | 39.4 | 67.9 | 80.9 | — | — | |
| FVNetwork=VGG2021.04 | 39.4 | 67.9 | 80.9 | — | — | |
| DVSA2015.11 | 38.4 | — | 80.5 | 1 | — | |
| DVSABackbone=R-CNN, AlexNet2019.09 | 38.4 | 69.9 | 80.5 | — | — | |
| DSVANetwork=AlexNet2021.04 | 38.4 | 69.9 | 80.5 | — | — | |
| m-CNN2015.11 | 38.3 | — | 81 | 2 | — | |
| m-CNNSupervision type=Direct supervision2017.05 | 38.3 | — | 81 | 2 | — | |
| SkipThoughtSupervision type=Pre-trained, Backbone=ResNet101, Training samples count=113k2017.05 | 37.9 | 72.2 | 84.3 | 2 | — | |
| Order2019.09 | 37.9 | — | 85.9 | — | — | |
| STV2015.11 | 33.8 | — | 82.1 | 3 | — | |
| SkipThoughtSupervision type=Pre-trained, Backbone=VGG19, Training samples count=82k2017.05 | 33.8 | 67.7 | 82.1 | 3 | — | |
| order-embeddingsVariant=reversed2015.11 | 11.2 | — | 44 | 14.2 | 86.6 |