EURASIP Journal on Image and Video Processing

Table 6 Prediction accuracy comparison on UT-Interaction (%)

From: Human activity prediction using saliency-aware motion enhancement and weighted LSTM network

Method	Accuracy with half videos	Accuracy with full videos
Hierarchical-M [19]	82.3	88.4
TGTW [17]	72.1	89.5
HSOM [13]	84.0	93.8
Dynamic BOW [15]	66.1	78.5
Multi-stage LSTM [18]	79.3	90.0
P-TS [21]	89.7	94.6
RU-LSTM [24]	91.4	95.3
WLSTM (two modalities)	92.6	96.7

Back to article page