Visual Question Answering on MemexQA (test)
79.5Accuracy (How Many)Soft Attention
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Soft AttentionAttention=Yes2018.06 | 79.5 | 69.7 | 34.6 | 60.4 | 58.2 | 62.1 | |
| DMN+Attention=Yes2018.06 | 79.2 | 61.6 | 34.6 | 24.8 | 22.4 | 48 | |
| Bi-directional Attention FlowAttention=Yes2018.06 | 79 | 68.9 | 35.6 | 56.7 | 46.8 | 59.8 | |
| Embedding + LSTM + ConcatAttention=None2018.06 | 77.6 | 66.8 | 39.8 | 43.3 | 40.9 | 56.3 | |
| Multimodal Compact Bilinear PoolingAttention=Yes2018.06 | 77.3 | 61.8 | 25 | 22.9 | 24.8 | 46.2 | |
| Embedding + LSTMAttention=None2018.06 | 77.1 | 56.4 | 34.9 | 31.4 | 31 | 47.8 | |
| TGIF Temporal AttentionAttention Type=Temporal Attention Only2018.06 | 76.1 | 70 | 52.2 | 58.2 | 47.7 | 63 | |
| FVTAAttention Type=Focal Visual-Text Attention2018.06 | 76.1 | 71.4 | 47.6 | 67.6 | 66.8 | 66.9 | |
| Logistic RegressionAttention=None2018.06 | 64.5 | 24.1 | 21.7 | 27.7 | 26 | 29.5 |