Table 3 Prediction accuracy comparison of different modalities on sub-JHMDB (%)
From: Human activity prediction using saliency-aware motion enhancement and weighted LSTM network
Modality | Accuracy with half videos | Accuracy with full videos |
|---|---|---|
RGB-SME | 55.4 | 71.1 |
OF-SME | 56.7 | 73.8 |
Two-modality fusion | 66.8 | 78.1 |