Table 4 Evaluation results of our model and existing main methods in terms of ADD(-S) on the Linemod dataset. Bold names are symmetric objects.

From: Enhancing object pose estimation for RGB images in cluttered scenes

 

PoseCNN5

Efficient pose24

RePose34

Trans6D39

Ullah et al.26

BDR6D22

OURS

Ape

77.0

87.71

79.5

88.3

87.78

90.6

86.09

Bench vi.

97.5

99.71

100

99.4

99.20

98.2

100

Camera

93.5

97.94

99.2

97.8

99.32

98.5

96.76

Can

96.5

98.52

99.8

99.1

98.63

99.1

99.01

Cat

82.1

98.00

97.9

93.2

98.22

97.6

99.40

Driller

95.0

99.90

99.0

99.5

99.30

98.6

99.80

Duck

77.7

90.99

80.3

87.8

92.00

94.5

92.11

Eggbox

97.1

100

100

100

100

99.9

100

Glue

99.4

100

98.3

99.8

99.96

98.3

100

Holepuncher

52.8

95.15

96.9

96.7

96.23

92.7

95.14

Iron

98.3

99.69

100

99.9

99.46

98.4

99.69

Lamp

97.5

100

99.8

99.7

99.86

97.3

100

Phone

87.7

97.98

98.9

99.5

99.20

94.8

98.94

Average

88.6

97.35

96.1

96.9

97.62

96.8

97.45

  1. Significant values are in [bold].