AUTOMATIC RECONSTRUCTION OF INDOOR ENVIRONMENTS FOR SHARING AR AND VR SPACES

Recently the amount of applications that exploit the augmented reality has seen an hyperbolic growth. The progress of technology allowed people to deal with powerful tools simply by using the phone. Computer vision represents one of these fields of application; its goal is to allow the user to perceive, classify and elaborate data from the external world by means of a camera. This project consists in the development of an application to let the user to reconstruct an indoor environment by taking photos of the objects in it. Although a lot of applications allow the object placement in an Augmented Reality environment, this particular system differs by the fact that the reconstructed objects placed in the scene are not premade, but are created at runtime by an external server which receives the photo as input and produces the correspondent 3D model as output. The system is based on a client-server architecture, where the client side is developed using the Unity Environment and the AR functionalities provided by Google ARCore, and the server side is based on a pipeline which extracts the current object from the scene, reconstructs it as a 3D model, refines the mesh imperfections and returns it to the client rotated according to the real position of the object inside the scene.