Automatic Aesthetics Evaluation of Robotic Dance Poses Based on Hierarchical Processing Network

Table 4

The comparison between the state-of-the-art approaches and our approach.

	The approach in [8]	The approach in [9]	The approach in [10]	Our approach
Information channel	Nonvisual and visual	Nonvisual and visual	Visual	Visual
Cognitive model based	No	No	Yes	Yes
Feature type	Hand-crafted features	Hand-crafted features	Hand-crafted features	Automatically extracted features
Aesthetic manner	Machine learning based method	Machine learning based method	Machine learning based method	Deep learning based method
Data acquisition environment	Simulation	Simulation	Simulation	Real
Highest correct ratio	81.8%	81.6%	81.6%	82.3%