A generative modelling technique for 3D reconstruction from a single 2D image

Abstract

3D Object Reconstruction is the task of predicting the 3D model of an object given a set of 2D images. In this paper, we propose an approach to solving this problem, given a single 2D image. We attempt to make use of several deep learning techniques. Our model consists of two parts. The first part generates multiple images having different viewpoints. We have included this part because reconstructing 3D object directly from a single 2D image is quite difficult, but the same task would be a lot easier given multiple images which capture different views of that same object. Also, predicting an image having a different viewpoint is much easier than predicting the whole 3D object, given an input image. The second part uses a network consisting of an Encoder, a Decoder (or Generator), and a Discriminator to predict the complete 3D voxel grid of the object. In this way, we achieve significant improvements in the results as compared to the existing techniques. © 2020 IEEE.