Semantic Room Wireframe Detection from a Single View
This addresses a problem for indoor scene reconstruction in computer vision, but it is incremental as it builds on existing room layout estimation and wireframe detection approaches.
The paper tackles the difficulty of reconstructing indoor surfaces with limited or repeated textures by proposing a Semantic Room Wireframe Detection task to predict a semantic wireframe from a single image, showing that SRW-Net handles complex room geometries better than previous methods and outperforms the baseline in non-semantic wireframe detection.
Reconstruction of indoor surfaces with limited texture information or with repeated textures, a situation common in walls and ceilings, may be difficult with a monocular Structure from Motion system. We propose a Semantic Room Wireframe Detection task to predict a Semantic Wireframe from a single perspective image. Such predictions may be used with shape priors to estimate the Room Layout and aid reconstruction. To train and test the proposed algorithm we create a new set of annotations from the simulated Structured3D dataset. We show qualitatively that the SRW-Net handles complex room geometries better than previous Room Layout Estimation algorithms while quantitatively out-performing the baseline in non-semantic Wireframe Detection.