Research on continuous jump action recognition of feet based on improved C3D

So far, the two-legged continuous jump test is in the manual test stage. To solve this problem, a test method based on computer vision is proposed to replace the manual test. In this paper, the C3D network is improved, and the input image is changed to 8 frames of 112 × 112 size. The BN layer is added after the convolution operation to reduce the gradient mass; the SGD is replaced by the Adam optimizer, and the activation function uses a Gaussian error linear unit. The improved network model has higher recognition accuracy than the original network. The proposed method embodies the integration of artificial intelligence and physical fitness test, and opens up new ideas for students’ physical fitness test. It provides a unified and objective judgment standard, which effectively reduces the