Fig. 6From: Human activity prediction using saliency-aware motion enhancement and weighted LSTM networkSome sampled frames from UT-Interaction and sub-JHMDBBack to article page