Semantic Room Wireframe Detection

This repo contains code to generate Semantic Room Wireframe (SRW) annotations from Structured3D and LSUN. It also holds the implementation of SRW-Net. See the preprint of our paper at arXiv for more details, which will appear at ICPR2022.

The model predicts the coordinates and labels of the lines and junctions of the frame of an indoor scene. Lines can be labeled: “Wall”, “Floor”, “Ceiling”, “Window” or “Door” Junctions are labeled/ “Proper” - Plane junctions that are visual or occluded by an object “False” - Plane junctions that are occluded by image frame or other planes

Citation

If you use it in your research, please cite ``` @misc{srw-net, doi = {10.48550/ARXIV.2206.00491}, url = {https://arxiv.org/abs/2206.00491},
author = {Gillsjö, David and Flood, Gabrielle and Åström, Kalle}, title = {Semantic Room Wireframe Detection from a Single View}, publisher = {arXiv}, year = {2022}, }