Understanding Indoor Scene: Spatial Layout Estimation, Scene Classification, and Object Detection

In this paper, we seek to understand scene from different viewpoints such as estimating the spatial layout of indoor scenes, detecting objects in the scene and making scene classification. In the previous work, every step has been done in a separate process so in this work we combine the three steps in one algorithm. To understand the indoor scene this requires interpreting the estimation of multiple scene elements and studying the relation between them. We propose a method that takes a single indoor image and then the relation between scene elements can be automatically learned and applied to totally understand the indoor scenes. Our proposed method consists of two steps: the first step is estimating the spatial layout of indoor scenes and this is done by extracting line segments and estimating the three orthogonal vanishing points. The orientation of the surfaces in the scene is defined by the vanishing points and so constructs the layout of the walls, ceiling, and floor. The second step is studying the relation between scene elements and this means that studying the relation between scene classification, objects in the scene, and layout estimation. We conduct extensive experiments to evaluate our proposed method performance on a single indoor image. We achieve superior accuracy results on two public datasets and provide a robust estimation of scene type, estimation of spatial layout and 3D objects.