Table 3 Quantitative comparison results on MPI-INF-3DHP using the 2D ground truth as input.

From: Origin centric and part based pose decomposition for 3D human pose estimation

Method

T

PCK\(\uparrow\)

AUC\(\uparrow\)

MPJPE↓

VideoPose3D25

81

86.0

51.9

84.0

PoseFormer11

9

88.6

56.4

77.1

MixSTE13

27

94.4

66.5

54.9

HDFormer36

96

98.7

72.9

37.2

STCFormer35

81

98.7

83.9

23.1

MotionAGFormer16

81

98.2

85.3

16.2

Ours

81

98.7

85.3

15.9

Back to article page