JGRCAN: A Visual Question Answering Co-Attention Network via Joint Grid-Region Features

Figure 6

The overall architecture of the co-attention layer.