Performance Comparison of Deep Learning (DL)-Based Tabular Models for Building Mapping Using High-Resolution Red, Green, and Blue Imagery and the Geographic Object-Based Image Analysis Framework
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
Identifying urban buildings in high-resolution RGB images presents challenges, mainly due to the absence of near-infrared bands in UAVs and Google Earth imagery and the diversity in building attributes. Deep learning (DL) methods, especially Convolutional Neural Networks (CNNs), are widely used for building extraction but are primarily pixel-based. Geographic Object-Based Image Analysis (GEOBIA) has emerged as an essential approach for high-resolution imagery. However, integrating GEOBIA with DL models presents challenges, including adapting DL models for irregular-shaped segments and effectively merging DL outputs with object-based features. Recent developments include tabular DL models that align well with GEOBIA. GEOBIA stores various features for image segments in a tabular format, yet the effectiveness of these tabular DL models for building extraction still needs to be explored. It also needs to clarify which features are crucial for distinguishing buildings from other land-cover types. Typically, GEOBIA employs shallow learning (SL) classifiers. Thus, this study evaluates SL and tabular DL classifiers for their ability to differentiate buildings from non-building features. Furthermore, these classifiers are assessed for their capacity to handle roof heterogeneity caused by sun exposure and roof materials. This study concludes that some SL classifiers perform similarly to their DL counterparts, and it identifies critical features for building extraction.
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) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| 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