Fig. 10From: Human activity prediction using saliency-aware motion enhancement and weighted LSTM networkPrediction accuracy comparison under different video observation ratio. a UT-Interaction. b Sub-JHMDBBack to article page