A single RGB image based 3D object reconstruction system

Easy and fast digitalization of real objects is especially useful when considering augmented reality (AR) and virtual reality (VR), as reconstructed objects allow a better interaction between the real and virtual worlds than using pre-made 3D CAD models. Thanks to the ubiquity of smartphones and to the spread of immersive VR devices, the AR and VR technologies are rapidly becoming popular. However, an affordable, robust and easy to use solution for object digitalization is still missing. This paper presents a reconstruction system that allows users to convert a single photo of a real object into a digital 3D asset. A smartphone is used to capture a snapshot of the object, whereas a secondary computing device performs the reconstruction process by exploiting a pipeline of three Deep Learning methods. Several experiments have been conducted in order to assess the accuracy and robustness of the system by using a standard metric for measuring the reconstruction accuracy (chamfer distance). The main outcomes show that the proposed system has a comparable accuracy with respect to the state-of-the-art methods for 3D object reconstruction.