Table 3 AP Scores using the HRFormer and ViTPose backbones

From: Neonatal pose estimation in the unaltered clinical environment with fusion of RGB, depth and IR images

Model

HRFormer-S

HRFormer-B

ViTPose-S

ViTPose-B

RGB

0.747 ± 0.112

0.760 ± 0.099

0.714 ± 0.098

0.693 ± 0.129

Depth

0.788 ± 0.097

0.795 ± 0.091

0.657 ± 0.143

0.671 ± 0.139

IR

0.764 ± 0.093

0.776 ± 0.088

0.657 ± 0.131

0.711 ± 0.108

EIF-RGB-D

0.764 ± 0.086

0.783 ± 0.077

0.663 ± 0.076

0.698 ± 0.073

EIF-RGB-IR

0.770 ± 0.089

0.795 ± 0.077

0.645 ± 0.096

0.682 ± 0.077

EIF-D-IR

0.788 ± 0.087

0.795 ± 0.081

0.662 ± 0.14

0.671 ± 0.14

EIF-RGB-D-IR

0.780 ± 0.090

0.777 ± 0.079

0.653 ± 0.082

0.690 ± 0.079

IIF-1

0.785 ± 0.096

0.794 ± 0.094

0.751 ± 0.105

0.714 ± 0.100

IIF-2

0.800 ± 0.096

0.809 ± 0.085

0.715 ± 0.096

0.753 ± 0.121

IIF-3

0.785 ± 0.098

0.804 ± 0.088

0.710 ± 0.110

0.746 ± 0.118

IIF-4

0.753 ± 0.103

0.759 ± 0.095

0.721 ± 0.093

0.742 ± 0.091

LIF-1

0.752 ± 0.103

0.752 ± 0.103

0.671 ± 0.096

0.676 ± 0.110

LIF-2

0.772 ± 0.097

0.772 ± 0.097

0.661 ± 0.092

0.674 ± 0.103

LIF-3

0.764 ± 0.105

0.761 ± 0.104

0.625 ± 0.106

0.643 ± 0.106

LIF-4

0.741 ± 0.107

0.744 ± 0.096

0.655 ± 0.093

0.679 ± 0.101

  1. S indicates small configuration, B indicates base configuration. The best performance for each backbone is given in bold.