Table 6 Performance comparison of occlusion handling strategies on cheerleading sequences.

From: Research on the construction of cheerleading technique evaluation and teaching system integrating deep visual recognition and cognitive feedback mechanism

Method

PCK@0.2 (Visible)

PCK@0.2 (Occluded)

Temporal consistency

FPS

Single-view baseline

91.3%

62.7%

0.72

42.1

Multi-view averaging

93.1%

74.2%

0.81

38.6

Graph neural network

92.8%

78.9%

0.85

31.2

Our transformer-based

94.2%

85.3%

0.91

35.4

Back to article page