Sentence Retrieval on MSCOCO 5k (test)
80.9R@1RCAR
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RCARWarm-up=true, Model Variant=All2023.03 | 80.9 | 96.9 | 98.9 | — | — | |
| RCARModel Variant=All2023.03 | 80.6 | 96.6 | 98.6 | — | — | |
| GPOWarm-up=true, Ensemble=true2023.03 | 80 | 97 | 99 | — | — | |
| SGRAFBackbone=ResNet-101 BUTD [5]2023.03 | 79.6 | 96.2 | 98.5 | — | — | |
| RCARWarm-up=true, Model Variant=I2T2023.03 | 79.3 | 96.5 | 98.8 | — | — | |
| RCARWarm-up=true, Model Variant=T2I2023.03 | 79.1 | 96.5 | 98.8 | — | — | |
| RCARModel Variant=I2T2023.03 | 78.5 | 95.9 | 98.5 | — | — | |
| GSMNBackbone=ResNet-101 BUTD [5]2023.03 | 78.4 | 96.4 | 98.6 | — | — | |
| GPOWarm-up=true2023.03 | 78.4 | 96.2 | 98.7 | — | — | |
| RCARModel Variant=T2I2023.03 | 78.2 | 96.3 | 98.4 | — | — | |
| SHANBackbone=ResNet-101 BUTD [5]2023.03 | 76.8 | 96.3 | 98.7 | — | — | |
| IMRAMBackbone=ResNet-101 BUTD [5]2023.03 | 76.7 | 95.6 | 98.5 | — | — | |
| VSRNBackbone=ResNet-101 BUTD [5]2023.03 | 76.2 | 94.8 | 98.2 | — | — | |
| CAANBackbone=ResNet-101 BUTD [5]2023.03 | 75.5 | 95.4 | 98.5 | — | — | |
| WCGLBackbone=ResNet-101 BUTD [5]2023.03 | 75.4 | 95.5 | 98.6 | — | — | |
| MMCABackbone=ResNet-101 BUTD [5]2023.03 | 74.8 | 95.6 | 97.7 | — | — | |
| SCANWarm-up=true2023.03 | 72.8 | 94.3 | 98 | — | — | |
| SCANBackbone=ResNet-101 BUTD [5]2023.03 | 72.7 | 94.8 | 98.4 | — | — | |
| ImageBERTtraining_mode=fine-tuned2020.01 | 66.4 | 89.8 | 94.4 | — | — | |
| UNITERtraining_mode=fine-tuned2020.01 | 63.3 | 87 | 93.1 | — | — | |
| Unicoder-VLtraining_mode=fine-tuned2020.01 | 62.3 | 87.1 | 92.8 | — | — | |
| RCARWarm-up=true, Model Variant=All2023.03 | 61.3 | 86.1 | 92.6 | — | — | |
| GPOWarm-up=true, Ensemble=true2023.03 | 59.8 | 86.1 | 92.8 | — | — | |
| RCARModel Variant=All2023.03 | 59.6 | 85.8 | 92.4 | — | — | |
| TERAN MrSw (ens.)Backbone=Region CNN, Ensemble=true2020.08 | 59.3 | 85.8 | 92.4 | 65.8 | 62.4 | |
| RCARWarm-up=true, Model Variant=T2I2023.03 | 59.1 | 84.8 | 91.8 | — | — | |
| RCARWarm-up=true, Model Variant=I2T2023.03 | 58.4 | 84.6 | 91.9 | — | — | |
| SGRAF2021.01 | 57.8 | — | 91.6 | — | — | |
| SGRAFBackbone=ResNet-101 BUTD [5]2023.03 | 57.8 | 84.9 | 91.6 | — | — | |
| RCARModel Variant=T2I2023.03 | 57.4 | 83.8 | 91 | — | — | |
| SGR2021.01 | 56.9 | — | 90.5 | — | — | |
| GPOWarm-up=true2023.03 | 56.8 | 84.5 | 91.4 | — | — | |
| RCARModel Variant=I2T2023.03 | 56.6 | 83.3 | 91.2 | — | — | |
| TERAN MrSwBackbone=Region CNN, Model Configuration=Image-Region / Sentence-Word2020.08 | 55.6 | 83.9 | 91.6 | 64.3 | 60.6 | |
| CAMERABackbone=Region CNN, Language Model=BERT, Pipeline Type=disentangled visual-textual, Ensemble=true2020.08 | 55.1 | 82.9 | 91.2 | — | — | |
| TERAN Symm.Backbone=Region CNN, Model Configuration=Symmetric matching2020.08 | 54.8 | 82.7 | 90.9 | 64.1 | 60.1 | |
| MMCA2021.01 | 54 | — | 90.7 | — | — | |
| MMCABackbone=Region CNN, Language Model=BERT2020.08 | 54 | 82.5 | 90.7 | — | — | |
| MMCABackbone=ResNet-101 BUTD [5]2023.03 | 54 | 82.5 | 90.7 | — | — | |
| IMRAM2021.01 | 53.7 | — | 91 | — | — | |
| Full-IMRAMBackbone=Region CNN2020.08 | 53.7 | 83.2 | 91 | — | — | |
| IMRAMBackbone=ResNet-101 BUTD [5]2023.03 | 53.7 | 83.2 | 91 | — | — | |
| SAF2021.01 | 53.3 | — | 90.1 | — | — | |
| CAMERABackbone=Region CNN, Language Model=BERT, Pipeline Type=disentangled visual-textual2020.08 | 53.1 | 81.3 | 89.8 | — | — | |
| VSRN2021.01 | 53 | — | 89.4 | — | — | |
| VSRNBackbone=Region CNN, Pipeline Type=disentangled visual-textual, Ensemble=true2020.08 | 53 | 81.1 | 89.4 | 65.2 | 61.2 | |
| VSRNBackbone=ResNet-101 BUTD [5]2023.03 | 53 | 81.1 | 89.4 | — | — | |
| CAAN2021.01 | 52.5 | — | 90.9 | — | — | |
| CAANBackbone=ResNet-101 BUTD [5]2023.03 | 52.5 | 83.3 | 90.9 | — | — | |
| AAMELBackbone=ResNet2020.08 | 51.9 | 84.2 | 91.2 | — | — | |
| SCANtraining_mode=fine-tuned2020.01 | 50.4 | 82.2 | 90 | — | — | |
| SCAN2021.01 | 50.4 | — | 90 | — | — | |
| SCANBackbone=Region CNN, Ensemble=true2020.08 | 50.4 | 82.2 | 90 | — | — | |
| SCANBackbone=ResNet-101 BUTD [5]2023.03 | 50.4 | 82.2 | 90 | — | — | |
| VSRN*Model Setup=single model2021.01 | 50.3 | — | 87.9 | — | — | |
| VSRNBackbone=Region CNN, Pipeline Type=disentangled visual-textual2020.08 | 50.3 | 79.6 | 87.9 | 63.9 | 59.8 | |
| CAMP2021.01 | 50.1 | — | 89.7 | — | — | |
| SCANWarm-up=true2023.03 | 50.1 | 79.5 | 88.1 | — | — | |
| SGM2021.01 | 50 | — | 87.9 | — | — | |
| M3A-NetBackbone=ResNet2020.08 | 48.9 | 75.2 | 84.4 | — | — | |
| TERAN MwSrBackbone=Region CNN, Model Configuration=Image-Word / Sentence-Region2020.08 | 45.3 | 76.3 | 86.2 | 61.1 | 56.4 | |
| VSE++Backbone=ResNet, Pipeline Type=disentangled visual-textual2020.08 | 41.3 | 71.1 | 81.2 | 59.7 | 55.1 | |
| TERN w. AlignBackbone=Region CNN, Alignment Strategy=Fine-grained alignments2020.08 | 40.2 | 71.1 | 81.9 | 60.6 | 56.1 | |
| TERNBackbone=Region CNN2020.08 | 38.4 | 69.5 | 81.3 | 60.1 | 55.6 | |
| TERN MwSi TestBackbone=Region CNN, Model Configuration=Image-Word / Sentence-Global2020.08 | 35.5 | 67.5 | 78.9 | 60 | 55.1 | |
| MRNNBackbone=Region CNN2020.08 | 16.5 | 39.2 | 52 | — | — | |
| TERN MiSw TestBackbone=Region CNN, Model Configuration=Image-Global / Sentence-Word2020.08 | 6.8 | 28.4 | 46.7 | 40.6 | 37.2 |