Fig. 8From: Human activity prediction using saliency-aware motion enhancement and weighted LSTM networkComparison of confusion matrices for three datasets (two modalities). a UT-Interaction set 1. b UT-Interaction set 2. c Sub-JHMDBBack to article page