Visual Experience-Based Question Answering with Complex Multimodal Environments

Table 7

Performance analysis of relationship recognition depending on different input information.

Input information	Precision (%)	Recall (%)	Accuracy (%)
3D Bbox	75.28	80.72	86.04
Object Class	31.82	58.24	67.53
3D Bbox + Object Class	76.34	94.24	89.67

Bold values represent the best results.