Visual Question Answering on VQA 1.0 (test-dev)
67.42Overall AccuracyMUTAN
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MUTANEnsemble size=52017.05 | 67.42 | 85.14 | 39.81 | 58.52 | |
| MUTANEnsemble size=32017.05 | 67.03 | 84.57 | 39.32 | 57.36 | |
| MLBEnsemble size=72017.05 | 66.77 | 84.54 | 39.21 | 57.81 | |
| MCBEnsemble size=72017.05 | 66.7 | 83.4 | 39.8 | 58.5 | |
| QGHC+Att+Concatfusing mechanism=attention2018.08 | 65.89 | 83.54 | 38.06 | 57.1 | |
| MFB+CoAttfusing mechanism=attention2018.08 | 65.1 | 83.2 | 38.8 | 55.5 | |
| MLB+Attfusing mechanism=attention2018.08 | 65.08 | 84.02 | 37.9 | 54.77 | |
| EENMNfusing mechanism=attention2018.08 | 64.9 | — | — | — | |
| End-to-End Module Networks (policy search after cloning)Visual feature=ResNet-1522017.04 | 64.9 | — | — | — | |
| MCBVisual feature=ResNet-1522017.04 | 64.7 | — | — | — | |
| MFB+Attfusing mechanism=attention2018.08 | 64.6 | 82.5 | 38.3 | 55.2 | |
| OursBackbone=ResNet2017.04 | 64.5 | 82.2 | 39.1 | 55.2 | |
| DANBackbone=ResNet2017.04 | 64.3 | 83 | 39.1 | 53.9 | |
| DANfusing mechanism=attention2018.08 | 64.3 | 83 | 39.1 | 53.9 | |
| MCBBackbone=ResNet2017.04 | 64.2 | 82.2 | 37.7 | 54.8 | |
| MCB+Attfusing mechanism=attention2018.08 | 64.2 | 82.2 | 37.7 | 54.8 | |
| End-to-End Module Networks (cloning expert)Visual feature=ResNet-1522017.04 | 64.2 | — | — | — | |
| RAUBackbone=ResNet2017.04 | 63.3 | 81.9 | 39 | 53 | |
| RAUfusing mechanism=attention2018.08 | 63.3 | 81.9 | 39 | 53 | |
| MODERNfusing mechanism=attention, backbone=ResNet-1522018.08 | 62.16 | 81.38 | 36.06 | 51.64 | |
| DANBackbone=VGG2017.04 | 62 | 82.1 | 38.2 | 50.2 | |
| End-to-End Module Networks (cloning expert)Visual feature=LRCN VGG-162017.04 | 61.9 | — | — | — | |
| HieCoAttBackbone=ResNet2017.04 | 61.8 | 79.7 | 38.7 | 51.7 | |
| MHieCoAttfusing mechanism=attention2018.08 | 61.8 | 79.7 | 38.7 | 51.7 | |
| HieCoAtt2017.05 | 61.8 | 79.7 | 38.7 | 51.7 | |
| MRNBackbone=ResNet2017.04 | 61.7 | 82.3 | 38.8 | 49.3 | |
| MRN2017.05 | 61.68 | 82.28 | 38.82 | 49.25 | |
| DMN+Backbone=VGG2017.04 | 60.3 | 80.5 | 36.8 | 48.3 | |
| MRNfusing mechanism=attention2018.08 | 59.4 | 80.81 | 35.91 | 46.43 | |
| DNMNfusing mechanism=attention2018.08 | 59.4 | 81.1 | 38.6 | 45.4 | |
| D-NMN2017.05 | 59.4 | 81.1 | 38.6 | 45.5 | |
| ACKBackbone=VGG2017.04 | 59.2 | 81 | 38.4 | 45.2 | |
| ACK2017.05 | 59.17 | 81.01 | 38.42 | 45.23 | |
| SANBackbone=VGG2017.04 | 58.7 | 79.3 | 36.6 | 46.1 | |
| SANfusing mechanism=attention2018.08 | 58.7 | 79.3 | 36.6 | 46.1 | |
| SAN2017.05 | 58.7 | 79.3 | 36.6 | 46.1 | |
| NMNBackbone=VGG2017.04 | 58.6 | 81.2 | 38 | 44 | |
| NMNfusing mechanism=attention2018.08 | 58.6 | 81.2 | 38 | 44 | |
| Ask Your Neur.2017.05 | 58.39 | 78.39 | 36.45 | 46.28 | |
| SMemfusing mechanism=attention2018.08 | 58 | 80.9 | 37.3 | 43.1 | |
| SMem 2-hop2017.05 | 57.99 | 80.87 | 37.32 | 43.12 | |
| D-NMNVisual feature=LRCN VGG-162017.04 | 57.9 | — | — | — | |
| VQA team2017.04 | 57.8 | 80.5 | 36.8 | 43.1 | |
| NMNVisual feature=LRCN VGG-162017.04 | 57.3 | — | — | — |