Fig. 2From: A novel Multi-Layer Attention Framework for visual description prediction using bidirectional LSTMBase Model Architecture [8]Back to article page