Uploaded August 2020 | Updated November 2026, 2 hours ago
Our new surface reconstruction method uses a fine local prior together with a coarse global one. With this, we can reconstruct surfaces more accurately than Screened Poisson Surface Reconstruction, DeepSDF and AtlasNet. Our method is independent from object classes and generalizes to unseen acquisition methods. On top, it doesn't need point normals.
This is the long video for ECCV 2020.
Publication: https://www.cg.tuwien.ac.at/research/publications/2020/erler-p2s/
Code and Datasets: github.com/ErlerPhilipp/points2surf
Our new surface reconstruction method uses a fine local prior together with a coarse global one. With this, we can reconstruct surfaces more accurately than Screened Poisson Surface Reconstruction, DeepSDF and AtlasNet. Our method is independent from object classes and generalizes to unseen acquisition methods. On top, it doesn't need point normals.
This is the long video for ECCV 2020.
Publication: https://www.cg.tuwien.ac.at/research/publications/2020/erler-p2s/
Code and Datasets: github.com/ErlerPhilipp/points2surf










