H2O-SDF: Two-phase Learning for 3D Indoor Reconstruction using Object Surface Fields
This addresses the problem of detailed 3D reconstruction in indoor environments for applications like robotics or VR, but it appears incremental as it builds on existing NeRF/SDF techniques.
The paper tackles 3D indoor scene reconstruction by introducing H2O-SDF, a two-phase learning method that uses an Object Surface Field to distinguish object and non-object regions, achieving a balance between preserving room layouts and capturing object details.
Advanced techniques using Neural Radiance Fields (NeRF), Signed Distance Fields (SDF), and Occupancy Fields have recently emerged as solutions for 3D indoor scene reconstruction. We introduce a novel two-phase learning approach, H2O-SDF, that discriminates between object and non-object regions within indoor environments. This method achieves a nuanced balance, carefully preserving the geometric integrity of room layouts while also capturing intricate surface details of specific objects. A cornerstone of our two-phase learning framework is the introduction of the Object Surface Field (OSF), a novel concept designed to mitigate the persistent vanishing gradient problem that has previously hindered the capture of high-frequency details in other methods. Our proposed approach is validated through several experiments that include ablation studies.