DuLa-Net: A Dual-Projection Network for Estimating Room Layouts from a Single RGB Panorama
This work addresses the challenge of room layout estimation for applications like virtual reality or robotics, but it is incremental as it builds on existing projection-based methods with a novel fusion approach.
The authors tackled the problem of estimating 3D room layouts from single RGB panoramas by proposing DuLa-Net, a dual-projection network that uses equirectangular and ceiling views, and introduced the Realtor360 dataset; the method outperformed state-of-the-art in accuracy, particularly for non-cuboid layouts.
We present a deep learning framework, called DuLa-Net, to predict Manhattan-world 3D room layouts from a single RGB panorama. To achieve better prediction accuracy, our method leverages two projections of the panorama at once, namely the equirectangular panorama-view and the perspective ceiling-view, that each contains different clues about the room layouts. Our network architecture consists of two encoder-decoder branches for analyzing each of the two views. In addition, a novel feature fusion structure is proposed to connect the two branches, which are then jointly trained to predict the 2D floor plans and layout heights. To learn more complex room layouts, we introduce the Realtor360 dataset that contains panoramas of Manhattan-world room layouts with different numbers of corners. Experimental results show that our work outperforms recent state-of-the-art in prediction accuracy and performance, especially in the rooms with non-cuboid layouts.