Generating Point Cloud from Measurements and Shapes Based on Convolutional Neural Network: An Application for Building 3D Human Model

Table 3

Average error on each part of 3D male and female models after activating the CNN model (the test set is comprised of damaged and undamaged samples).

Part	Train/male (%)	Train/female (%)	Test/male (undamaged) (%)	Test/female (undamaged) (%)	Test/male (damaged) (%)	Test/female (damaged) (%)	Test/male (average) (%)	Test/female (average) (%)
Torso	7.42	11.26	6.46	9.74	10.73	15.79	8.59	12.765
Left arm	8.24	15.20	6.68	15.83	10.87	26.11	8.77	20.97
Right arm	12.99	12.23	10.83	12.33	14.72	13.82	12.77	13.075
Left leg	8.78	7.26	7.59	7.40	11.22	9.92	9.40	8.66
Right leg	8.39	8.01	7.81	7.45	11.81	13.26	9.81	10.35
Average	9.16	10.59	7.87	10.55	11.87	15.78	9.86	13.16