Skip to main content

Table 1 Comparison of different methods using the COCO validation set

From: Classroom student posture recognition based on an improved high-resolution network

Method

Input size

#Params

GFLOPs

AP

AR

OpenPose [7]

368 × 368

–

–

61.8

66.5

Baseline ResNet-50 [28]

256 × 192

34.0 M

8.90

70.4

76.3

HRNet-W32 (paper) [11]

256 × 192

28.5 M

7.10

73.4

78.9

HRNet-W32 (our implement)

256 × 192

28.54 M

7.20

73.1

78.7

SE-HRNet-W32 (our)

256 × 192

28.75 M

7.21

73.8

79.2