Room layout estimation in indoor environment: a review

Scene understanding from the single image of an indoor scene is identified as a challenging task. This involves interpreting the assessment of multiple scene components, such as to identify the spatial layout from room, detect the objects in 3D space and classify the scene, to understand the nature of an indoor scene. Assessing the spatial structure of indoor scenes offers important geometric details and limits for various activities, such as indoor 3D-reconstruction, navigation, scene awareness and virtual reality. Most of the layout estimation function lacks the clutter and decorations present in the background, instead concentrating mainly on the horizontal wall contours. Room layout states the orientations, heights, and positions of walls with respective to its camera center. Then, a set of estimated boundaries or corner positions, or as a 3D mesh is characterized from layout. Nevertheless, identifying the 3D model from a single 2D picture is an ill-posed and challenging problem. Large number of techniques applies "Manhattan assumption" for layout estimation. The layout estimation approach is combined with few computer vision based techniques like semantic segmentation, object orientation estimation, scene classification, and 3D object detection to achieve indoor scene understanding. Additionally, the object detection and semantic segmentation is also reviewed along with this layout estimation approach. This survey provides valuable information to all researchers and those who are looking for better techniques in layout estimation.