Reconstruction of Building LoD2 Wireframe Models Using Semantic Segmentation
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Abstract. LoD2 building models can be used in different digital twin-related applications such as urban planning, disaster management, optimizing green energy efficiency, and solar panel recommendation. Existing technology for 3D modelling of buildings still relies on a large amount of manual work due to the irregular geometries of different roof types. Wireframes have shown to be an effective representation for 3D building especially in LoD2 format. Due to the complexity and diversity of roof types in urban areas, 3D building modeling remains a challenging task. In this paper, we propose a new framework for generating 3D wireframes to model different roof types. While high-resolution airborne images can be utilized to exploit the fine details of the roofs, they have difficulties in areas with poor contrast or shadows. The proposed framework incorporates the Digital Surface Model (DSM) as an auxiliary data source to address this limitation. In this work, we focus on the extraction of roof geometrical components including lines and planes of individual buildings to achieve a consistent LoD-2 building reconstruction. The proposed methodology is divided into two phases: (1) jointly predicting building lines and roof planes from the RGB imagery and DSM and (2) generating 3D wireframes of buildings using the extracted roof planes and lines. Subsequently, height values from the point clouds are used to derive 3D wireframes. Experiments with 1,620 buildings from Fredericton, the capital of New Brunswick in eastern Canada, demonstrate an IoU of 0.9337, an F1-score of 0.939, and an F2-score of 0.9378 for the roof geometrical components detection phase, as well as an RMSE of around 0.2-0.8 meter for the final 3D building model compared to the original LiDAR data was achieved.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it