Picture to 3D scene generation framework creates a conceivable 3D graphical scene from a given 2D advanced picture input. Because of overpowering utilization of 3D models in computer games, robot routes, virtual situations and even in field of interior designing there is a developing enthusiasm for 3D scene generation, scene comprehension and 3D demonstrate recovery. Here proposed another framework which is the mix of profound learning with spatial information portrayal to construct a rearranged 3D demonstrate from a solitary picture. The info picture is changed over into characteristic language depiction by utilizing a neural system. At last an improved 3D scene created from this depiction utilizing semantic parsing and spatial learning. The framework is extremely compelling at delivering important 3D scene straight forwardly from a 2D picture. The new framework additionally presented a system for rendering and controlling the scene through iterative info directions.