Research on Methods of Physical Aided Education Based on Deep Learning

Table 1

Structural parameters of the optimization model.

Training data set	Human images in the COCO data set and acquiring the badminton player images
Input number of images	10
Maximum iteration	200000
Basic learning rate	5 × 10^−5
Initial value of weight value	5 × 10^−4
Initial value of weight correction	0.01
The optimizer	Stochastic gradient descent
	Loss function	Pafmap human bone point grayscale of the preset image
Heatmap human bone point grayscale of the preset image