Objective The reliability and validity of the SGGC-Net-based motion capture system (hereafter referred to as the SGGC-Net system) and the SIMI system for parsing walking gait were compared using a Marker-point 3D motion capture system (Vicon) as a reference standard. Methods Thirty healthy college students were recruited and walked on a treadmill. Kinematic data were collected using the Vicon system and video data were collected synchronously using 4 cameras. Their right shoulder, elbow, hip, knee, and ankle joint angles were obtained. Reliability was assessed using intra-class correlation coefficients with their 95% confidence intervals and standard errors of measurement; validity was assessed using multiple correlation coefficients and root mean square errors. Results The intra-class correlation coefficients of the maximum and minimum angles of the three-dimensional coordinate angles of the upper and lower limb joints of the SGGC-Net system ranged from 0.798-0.990, with the standard error of measurement of 0.04-0.95°,and the intra-class correlation coefficients of the SIMI system ranged from 0.650-0.967, with the standard error of measurement of 0.31-1.24°. The intra-class correlation coefficients of the SGGC-Net system were higher than those in the SIMI system for all the joint angles except for the minimum hip angle and maximum knee angle. Compared with the joint angle curves derived from the SIMI system, the multiple correlation coefficients of the curves derived from the SGGC-Net system ranged from 0.945-0.996, with the root-mean-square error of 1.44-4.65°, and the multiple correlation coefficients of the SIMI system ranged from 0.815-0.986, with the root-mean-square error of 2.56-9.99°. The multiple correlation coefficients of the SGGC-Net system were greater than those of the SIMI system in all angles except the ankle joint; the root mean square errors of the SGGC-Net system were smaller than those of the SIMI system in all angles except the ankle joints. Conclusions The SGGC-Net system has better reliability and validity than the SIMI system in most of the variables, and it has better repeatability and accuracy in analyzing the walking gait, which can be applied to motion capture environments without Marker points, such as technical analysis of athletes' movements and clinical gait analysis of special populations.