Table 2 Performance comparison on the MS COCO 2017 dataset.

From: A human pose estimation network based on YOLOv8 framework with efficient multi-scale receptive field and expanded feature pyramid network

Method

Backbone

Parameter (M)

AP50

HigherHRNet21

HRNet-w32

28.6

87.5

DEKR22

HRNet-w32

29.6

87.7

YOLO-Pose24

CSPDarkNet

41.4

89.1

YOLOv5m6-Rlepose27

CSPDarkNet

39.6

89.2

RTMO-m28

CSPDarkNet

22.6

89.0

YOLOv8m-Pose

CSPDarkNet

26.4

88.8

YOLOv8 + EFPN

CSPDarkNet

19.5

88.1

YOLOv8 + EFPN + Wise-IoU

CSPDarkNet

19.5

88.4

EE-YOLOv8

CSPDarkNet

19.6

89.0