3DVAEReCNN: Region-based Convolutional Neural Network for Volumetric Rendering of Indoor Scenes

3D Reconstructions are being appreciated across various fields as a more informative means of visualisation that offers great insight about the qualitative characteristics of the objects or scene under consideration. Hence more research is being carried out in this area as 3D are proving to be of great help to fields such as medicine, for improving the diagnostic accuracy and surgical precision of the medical procedure. It has also found applications in fields such as intelligent robot navigation by reproduction of the depth map of a scene, object recognition and so on. We present a unique proposition to synthesize volumetric reconstructions from a singular or multiple positions of view, based on RGB images/videos. The project aims at rapidly generating all the different parts of the indoor environment, without having to actually observe them in reality.