Visual Experience-Based Question Answering with Complex Multimodal Environments

Table 1

Specification of the VEQA dataset.

Category		Count
Action scenario	Action scenarios	200
	Actions per action scenario	77
Question	Existence	1,168
	Counting	1,168
	Attribute	1,168
	Relation	1,005
	Include	676
	AgentHas	212
	Total questions	5,397
	Vocabulary size	90
Scene graph	Scene graphs	3,916
	Objects	13,109
	Attributes	26,218
	Relationships	25,583