Human Action Recognition Algorithm Based on Improved ResNet and Skeletal Keypoints in Single Image

Table 3

Comparison of experimental results on different network models.

Method	RCNN	Action mask	RCNN	Whole and Parts	TDP	Ours
Jumping	91.3	86.7	91.5	84.7	96.4	93.6
Phoning	77.1	72.2	84.4	67.8	84.7	90.5
Playing instrument	91.1	94	93.6	91	96.7	96.8
Reading	76.1	71.3	83.2	66.6	83.3	90.5
Riding horse	96.7	95.4	96.9	96.6	99.4	98.4
Riding bike	96.3	97.6	98.4	97.2	99.2	98.6
Running	92	88.5	93.8	90.2	91.9	94.2
Taking photo	83.8	72.4	85.9	76	85.3	90.1
Using computer	85.9	88.4	92.6	83.4	93.9	94.5
Walking	81.8	65.3	81.8	71.6	84.7	85.2
mAP	87	83.2	90.2	82.6	91.6	92.1