A Two-Stage Diffusion Framework for 3D Building Reconstruction from Single Nadir Images
| dc.contributor.advisor | Sohn, Gunho | |
| dc.contributor.author | Korny, Youssef Shaaban Mohammed | |
| dc.date.accessioned | 2026-11-06T16:16:28Z | |
| dc.date.available | 2026-11-06T16:16:28Z | |
| dc.date.copyright | 2026-06-11 | |
| dc.date.issued | 2026-11-06 | |
| dc.date.updated | 2026-11-06T16:16:27Z | |
| dc.degree.discipline | Earth & Space Science | |
| dc.degree.level | Master's | |
| dc.degree.name | MSc - Master of Science | |
| dc.description.abstract | Reconstructing 3D urban geometry from single nadir aerial images is an ill-posed problem due to severe vertical ambiguity and occluded building facades. To resolve this, this thesis proposes GeoPriorPC, a novel two-stage generative framework bridging foundational 2D image priors and unstructured 3D point clouds. Stage 1 fine-tunes a Latent Diffusion Model (LDM) to extract a compressed geometric prior, specifically, oblique surface normals, from flat nadir inputs. Stage 2 employs a Diffusion Point Transformer (DiPT) backbone. It utilizes a dual-conditioning mechanism, integrating global adaptive Layer Normalization (adaLN-Zero) and local spatial cross-attention, to structurally guide the 3D denoising process using the learned 2D prior. Evaluated on the AIM2PC benchmark, GeoPriorPC establishes a new state-of-the-art, achieving a 44% relative improvement in F1-score and reducing Earth Mover's Distance by 49%. By explicitly resolving depth ambiguity, this framework successfully synthesizes the accurate architectural manifolds critical for urban digital twins. | |
| dc.identifier.uri | https://hdl.handle.net/10315/44336 | |
| dc.language | en | |
| dc.rights | Author owns copyright, except where explicitly noted. Please contact the author directly with licensing requests. | |
| dc.subject | Artificial intelligence | |
| dc.subject | Remote sensing | |
| dc.subject.keywords | 3D building reconstruction | |
| dc.subject.keywords | Generative diffusion models | |
| dc.subject.keywords | Point cloud generation | |
| dc.subject.keywords | Nadir aerial imagery | |
| dc.subject.keywords | Remote sensing | |
| dc.subject.keywords | Latent Diffusion Models (LDM) | |
| dc.subject.keywords | Geometric priors | |
| dc.subject.keywords | Surface normal estimation | |
| dc.subject.keywords | Deep learning | |
| dc.subject.keywords | Computer vision | |
| dc.title | A Two-Stage Diffusion Framework for 3D Building Reconstruction from Single Nadir Images | |
| dc.type | Electronic Thesis or Dissertation |
Files
Original bundle
1 - 1 of 1
Loading...
- Name:
- Korny_Youssef_Shaaban_Mohammed_2026_MSc.pdf
- Size:
- 36.22 MB
- Format:
- Adobe Portable Document Format