An Intelligent System for 50-Meter Running Test Based on Image Recognition Technology

Aiming at solving the problems involving heavy tasks, difficult organization, and error-prone in 50-meter running test, we use the GhostNet model and FaceNet algorithm to detect and recognize the human body and face with the real-time acquisition of 50-meter running test process image, completing identification and calculation of the score. It can accomplish the intelligentization of 50-meter running test. The algorithm introduces a depthwise separable convolution module, which improves the inference speed. Experiments show that the algorithm achieves 16.7 FPS on GPU devices. The gap with the average score which calculated by people is just 0.16 seconds, meeting the requirements of the test system for fast and high accuracy.