Fig. 4From: Human activity prediction using saliency-aware motion enhancement and weighted LSTM networkStructure of an LSTM unitBack to article page