Reconstruction of a realistic 3D human model from a single 2D image

Generation of human 3D models has become one of the most interesting topics in the domain of computer vision. This is because the process of reconstructing 3D human body models from 2D images has of course attracted a lot of attention in the present days because of the potential importance of the same in several different application areas. Such are the virtual trial rooms for online shopping clothes, virtual reality for general and sporting activities and diagnostic assessment of neurological diseases. The ever increasing relevance of 3D models in fields like e-commerce underlines the requirements of converting 2D images into 3D reconstructions. In this study, a new method is presented to animating various human models from single images employing advanced mesh deformer and volume renderer. The effectiveness of the proposed method is tested on the THuman4.0 and DeepCap datasets, where the evaluation indicators include PSNR, SSIM, LPIPS, and FID. The proposed approach is shown to be more effective than the current state of the art methods, enhancing the existing methods of 3D human model reconstruction.