Visual Question Answering (VQA) on VQA v1 test-std
Metric: Accuracy (higher is better)
Metric: Accuracy (higher is better)
| # | Model↕ | Accuracy▼ | Extra Data | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | SAAA (ResNet) | 64.6 | No | Show, Ask, Attend, and Answer: A Strong Baseline... | 2017-04-11 | Code |
| 2 | RAU (ResNet) | 63.2 | No | Training Recurrent Answering Units with Joint Lo... | 2016-06-12 | - |
| 3 | HieCoAtt (ResNet) | 62.1 | No | Hierarchical Question-Image Co-Attention for Vis... | 2016-05-31 | Code |
| 4 | DMN+ | 60.4 | No | Dynamic Memory Networks for Visual and Textual Q... | 2016-03-04 | Code |
| 5 | SAN (VGG) | 58.9 | No | Stacked Attention Networks for Image Question An... | 2015-11-07 | Code |
| 6 | NMN+LSTM+FT | 58.7 | No | Neural Module Networks | 2015-11-09 | Code |