Open-Ended Visual Question Answering on VQA (test-dev)
81.1Yes/No AccuracyFDA
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| FDABackbone=ResNet2016.05 | 81.1 | 36.2 | 45.8 | 59.2 | |
| SMemBackbone=GoogLeNet2016.05 | 80.9 | 37.3 | 43.1 | 58 | |
| LSTM Q+IBackbone=VGGNet2016.05 | 80.5 | 36.8 | 43 | 57.8 | |
| DMN+Backbone=VGGNet2016.05 | 80.5 | 36.8 | 48.3 | 60.3 | |
| Hierarchical Co-Attention (Alternating) + ResNetBackbone=ResNet, Attention Mechanism=Alternating co-attention2016.05 | 79.7 | 38.7 | 51.7 | 61.8 | |
| Hierarchical Co-Attention (Alternating) + VGGBackbone=VGGNet, Attention Mechanism=Alternating co-attention2016.05 | 79.6 | 38.4 | 49.1 | 60.5 | |
| Hierarchical Co-Attention (Parallel) + VGGBackbone=VGGNet, Attention Mechanism=Parallel co-attention2016.05 | 79.5 | 38.7 | 48.3 | 60.1 | |
| SANBackbone=VGGNet2016.05 | 79.3 | 36.6 | 46.1 | 58.7 |