Video Question Answering on TGIF-QA v2 (test)
68.2Action AccCo-memory
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Co-memoryDynamic Fact Ensemble=true, Memory Update Cycles (T)=22018.03 | 68.2 | 74.3 | 51.5 | 4.1 | |
| Co-memoryDynamic Fact Ensemble=false, Memory Update Cycles (T)=22018.03 | 66.8 | 73.2 | 51 | 4.21 | |
| ST-TPFeatures=ResNet-152 + Flow CNN, Attention=Temporal2018.03 | 62.9 | 69.4 | 49.5 | 4.32 | |
| ST-TPFeatures=ResNet-152 + C3D, Attention=Temporal2018.03 | 60.8 | 67.1 | 49.3 | 4.4 | |
| STFeatures=ResNet-152 + C3D2018.03 | 60.1 | 65.7 | 48.2 | 4.38 | |
| VQA-MCBAggregation=aggr2018.03 | 58.9 | 24.3 | 25.7 | 5.17 | |
| ST-SPFeatures=ResNet-152 + C3D, Attention=Spatial2018.03 | 57.3 | 63.7 | 45.5 | 4.28 | |
| ST-SP-TPFeatures=ResNet-152 + C3D, Attention=Spatial + Temporal2018.03 | 57 | 59.6 | 47.8 | 4.56 | |
| Yu et al.2018.03 | 56.1 | 64 | 39.6 | 5.13 | |
| VIS+LSTMAggregation=avg2018.03 | 48.8 | 34.8 | 35 | 4.8 | |
| VIS+LSTMAggregation=aggr2018.03 | 46.8 | 56.9 | 34.6 | 5.09 | |
| VQA-MCBAggregation=avg2018.03 | 29.1 | 33 | 15.5 | 5.54 |