Fig. 3From: A novel Multi-Layer Attention Framework for visual description prediction using bidirectional LSTMStacked LSTM Model ArchitectureBack to article page