A Novel Approach of Intelligent Computing for Multiperson Pose Estimation with Deep High Spatial Resolution and Multiscale Features

Table 4

Comparison experiments on the COCO test2017 dataset. Results of compared methods are cited from [15].

Method	Backbone	Input size	AP	AP50	AP75	APm	APl	AR
G-RMI [38]	ResNet-50		64.9	85.5	71.3	62.3	70.0	69.7
CPN	ResNet-Inception		72.1	91.4	80.0	68.7	77.2	78.5
FAIR [37]	ResNeXt-101-FPN	—	69.2	90.4	77.0	64.9	76.3	75.2
G-RMI	ResNet-152		71.0	87.9	77.7	69.0	75.2	70.6
oks [37]	—	—	72.0	90.3	79.7	67.6	78.4	77.2
Bangbanggren [37]	ResNet-101	—	72.8	89.4	79.6	68.6	80.0	78.7
CPN	ResNet-Inception		73.0	91.7	80.9	69.5	78.1	79.0
SimpleBaseline	ResNet-152		73.7	91.9	81.1	70.3	80.0	79.0
Ours	HSRNet-59		72.6	91.2	79.9	69.0	78.9	77.8
Ours	HSR-MSNet-59		72.8	91.2	80.5	69.6	79.0	78.3