Table 7 2D Directed Masking Technique

From: Leveraging two-dimensional pre-trained vision transformers for three-dimensional model generation via masked autoencoders

Ratio %

Visible

ScanObjectNN

ModelNet40

Object-BG

Masking

81

Significant

94.0

93.63

89.80

Yes

80

Significant

93.89

93.33

88.70

Yes

-

Arbitrarily

93.60

93.03

88.53

Non-Significant

78

Significant

93.40

92.81

88.09

Yes

70

Significant

93.15

92.57

87.94

Yes

85

Significant

92.95

92.43

87.83

Yes

Back to article page