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

Figure 1

The overall architecture of JGRCAN.