Visual Experience-Based Question Answering with Complex Multimodal Environments

Table 3

Performance analysis of scene graph generation depending on different state estimation methods.

State estimation methods	Object mAP (%)	SGGen (%)
		Attribute	Relation	Total
VQAS without SP	63.32	56.62	40.99	50.11
VQAS	63.91	58.69	41.48	51.26

Bold values represent the best results.