Parametric 3D Visualization Modeling of the Human Body Based on 2D Photos

In this paper, a method for rapid visualization of parametric 3D modeling based on two 2D photographs is proposed. Firstly, the background difference method is used to detect and separate the human contours. Secondly, by preprocessing the SCAPE model and using deep learning to establish the correspondence between the contour map and the 3D human model, the improved GoogleNet network model is trained, and the corresponding size of the 3D human model can be reconstructed by inputting the actual human contour. Finally, the error analysis by setting the error evaluation function yields Ea values of 8.1% and 5.9% for the front view and side view, respectively, which achieves an advantage in the study with which the same work was performed.