Table 1.

Evaluation of pose estimation methods on AFLW and PoseTrack datasets.

Model	AFLW dataset				PoseTrack dataset
	Accuracy	Recall	F1 score	AUC	Accuracy	Recall	F1 score	AUC
HRNet (41)	91.36 ± 0.02	89.22 ± 0.03	90.15 ± 0.02	88.45 ± 0.03	90.78 ± 0.02	87.94 ± 0.02	89.27 ± 0.03	87.63 ± 0.02
SimpleBaseline (42)	88.95 ± 0.03	87.30 ± 0.02	86.45 ± 0.03	85.78 ± 0.03	89.60 ± 0.02	88.42 ± 0.03	87.12 ± 0.02	86.89 ± 0.03
DarkPose (43)	92.12 ± 0.03	91.55 ± 0.02	90.67 ± 0.02	89.73 ± 0.02	91.94 ± 0.02	89.71 ± 0.03	90.12 ± 0.03	89.30 ± 0.02
OpenPose (44)	89.88 ± 0.02	88.66 ± 0.02	87.75 ± 0.03	86.12 ± 0.03	88.49 ± 0.02	87.20 ± 0.03	86.45 ± 0.02	85.72 ± 0.03
DEKR (45)	93.22 ± 0.03	91.03 ± 0.03	90.89 ± 0.02	90.12 ± 0.03	92.11 ± 0.03	90.85 ± 0.02	89.50 ± 0.02	89.95 ± 0.03
PRTR (46)	90.45 ± 0.02	89.18 ± 0.03	88.24 ± 0.02	87.49 ± 0.02	91.34 ± 0.03	89.72 ± 0.02	88.60 ± 0.03	87.90 ± 0.02
Ours	94.56 ± 0.02	92.34 ± 0.02	91.89 ± 0.03	92.14 ± 0.03	94.10 ± 0.03	92.87 ± 0.02	91.45 ± 0.03	91.72 ± 0.02

The values in bold are the best values.