HorizonNet
HorizonNet implements a novel 1D representation approach for extracting room layouts, developed as a pure Python project. It supports inference of both cuboid and general-shaped room layouts from images, offering features like 3D layout viewing and precise rotation correction. The integration of pano stretch augmentation enhances its adaptability for various applications. HorizonNet includes extensive evaluation metrics, supports custom dataset preparation, and is compatible with multiple datasets, making it a comprehensive solution for 3D room layout analysis.
