Visual Experience-Based Question Answering with Complex Multimodal Environments

Table 6

Performance analysis of attribute recognition depending on different input information.

Input information	Output	Precision (%)	Recall (%)	Accuracy (%)
Image + 2D Bbox	Color	89.23	79.76	87.52
	Openness	82.24	33.33	82.24
Object Class	Color	53.01	43.39	49.99
	Openness	72.03	71.89	92.42
Image + 2D Bbox + Object Class	Color	89.71	80.55	89.87
	Openness	79.57	79.48	94.57

Bold values represent the best results.