Visual Question Answering on VQA 1 (test-standard)
83.3VQA Open-Ended Accuracy (All)Human
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Human2015.11 | 83.3 | 95.77 | 83.39 | 72.67 | — | — | — | — | — | — | — | — | |
| QGHC+Att+Concatfusing mechanism=attention2018.08 | 65.9 | — | — | — | — | — | — | — | — | — | — | — | |
| MLB+Attfusing mechanism=attention2018.08 | 65.07 | — | — | — | — | — | — | — | — | — | — | — | |
| DANfusing mechanism=attention2018.08 | 64.2 | — | — | — | — | — | — | — | — | — | — | — | |
| RAUfusing mechanism=attention2018.08 | 63.2 | — | — | — | — | — | — | — | — | — | — | — | |
| MHieCoAttfusing mechanism=attention2018.08 | 62.1 | — | — | — | — | — | — | — | — | — | — | — | |
| FDA2016.04 | 59.54 | — | — | — | — | — | — | — | — | — | — | — | |
| DNMNfusing mechanism=attention2018.08 | 59.4 | — | — | — | — | — | — | — | — | — | — | — | |
| SANfusing mechanism=attention2018.08 | 58.9 | — | — | — | — | — | — | — | — | — | — | — | |
| SAN(2, CNN)layers=2, backbone=CNN2015.11 | 58.9 | — | — | — | — | — | — | — | — | — | — | — | |
| SAN2016.04 | 58.9 | — | — | — | — | — | — | — | — | — | — | — | |
| NMNfusing mechanism=attention2018.08 | 58.7 | — | — | — | — | — | — | — | — | — | — | — | |
| SMem-VQA Two-Hop2015.11 | 58.24 | 80.8 | 37.53 | 43.48 | — | — | — | — | — | — | — | — | |
| SMemfusing mechanism=attention2018.08 | 58.2 | — | — | — | — | — | — | — | — | — | — | — | |
| D-LSTMLSTM layers=22016.04 | 58.16 | — | — | — | — | — | — | — | — | — | — | — | |
| D-NMN2016.04 | 58 | — | — | — | — | — | — | — | — | — | — | — | |
| MRNfusing mechanism=attention2018.08 | 57.39 | — | — | — | — | — | — | — | — | — | — | — | |
| DPPnet2015.11 | 57.36 | 80.28 | 36.92 | 42.24 | 62.69 | 80.35 | 38.79 | 52.79 | — | — | — | — | |
| DPPnetexternal training data=true2015.11 | 57.36 | 80.28 | 36.92 | 42.24 | — | — | — | — | — | — | — | — | |
| DPPnet2016.04 | 57.36 | — | — | — | — | — | — | — | — | — | — | — | |
| ACKexternal training data=true2015.11 | 55.98 | 79.05 | 36.1 | 40.61 | — | — | — | — | — | — | — | — | |
| ACK2016.04 | 55.98 | — | — | — | — | — | — | — | — | — | — | — | |
| iBOWIMG2015.11 | 55.89 | 76.76 | 34.98 | 42.62 | — | — | — | — | — | — | — | — | |
| iBOWIMG2016.04 | 55.89 | — | — | — | — | — | — | — | — | — | — | — | |
| LSTM Q+Iarchitecture=LSTM, input=Question + Image2015.11 | 54.1 | — | — | — | — | — | — | — | — | — | — | — | |
| LSTM Q+I2015.11 | 54.06 | — | — | — | — | — | — | — | — | — | — | — | |
| LSTM Q+I2015.11 | 54.06 | — | — | — | — | — | — | — | — | — | — | — | |
| LSTM Q+IInput=Question and Image, LSTM layers=12016.04 | 54.06 | — | — | — | — | — | — | — | — | — | — | — | |
| A+C+S-K-LSTM2016.03 | — | — | — | — | — | — | — | — | 59.5 | 81.1 | 45.9 | 37.18 | |
| D-LSTMLayers=22016.04 | — | — | — | — | — | — | — | — | 63.09 | — | — | — | |
| DDPnet2016.03 | — | — | — | — | — | — | — | — | 57.36 | 80.28 | 42.24 | 36.92 | |
| DNMN2016.03 | — | — | — | — | — | — | — | — | 59.44 | 80.98 | 45.81 | 37.48 | |
| DPPnet2016.04 | — | — | — | — | — | — | — | — | 62.69 | — | — | — | |
| FDA2016.04 | — | — | — | — | — | — | — | — | 64.18 | — | — | — | |
| Human2016.03 | — | — | — | — | — | — | — | — | 83.3 | 95.77 | 72.67 | 83.39 | |
| iBOWIMG2016.04 | — | — | — | — | — | — | — | — | 61.97 | — | — | — | |
| IBOWING2016.03 | — | — | — | — | — | — | — | — | 55.89 | 76.76 | 42.62 | 34.98 | |
| LSTM QInput=Question only2016.03 | — | — | — | — | — | — | — | — | 48.89 | 78.12 | 26.99 | 34.94 | |
| LSTM Q+I2016.04 | — | — | — | — | — | — | — | — | 57.57 | — | — | — | |
| LSTM Q+IInput=Question + Image2016.03 | — | — | — | — | — | — | — | — | 54.06 | 79.01 | 36.8 | 35.55 | |
| NMN2016.03 | — | — | — | — | — | — | — | — | 58.66 | 81.16 | 44.01 | 37.7 | |
| SAN2016.03 | — | — | — | — | — | — | — | — | 58.85 | 79.11 | 46.42 | 36.41 | |
| SMem2016.03 | — | — | — | — | — | — | — | — | 58.24 | 80.8 | 43.48 | 37.53 |