PolyAttractNet: Graph-Based Polygonal Segmentation of Building Footprints Using Attraction Field Maps
Bibliographic record
Abstract
Abstract. Since the launch of Landsat-1 in 1972, Earth observation satellites have undergone significant advancements, enabling the collection of vast amounts of high-resolution imagery. These satellites continuously provide critical data for monitoring urban expansion, infrastructure development, and disaster response. In recent years, the number of remote sensing satellites in orbit has increased substantially, generating extensive visual datasets essential for precise spatial mapping across civil, public, and military applications. One of the key challenges in utilizing satellite imagery is the automated reconstruction of building footprints, which demands high precision to account for variations in architectural styles. Traditional methods rely on manual or semi-automated approaches, which are often time-consuming and prone to inaccuracies. To address these limitations, this paper introduces PolyAttractNet, a novel deep learning framework designed to improve building boundary delineation in satellite imagery. Our approach incorporates Attraction Field Maps (AFMs) within a Graph Neural Network (GNN) framework, combined with an enhanced Mask R-CNN backbone. The proposed architecture effectively detects building instances from a single satellite image while minimizing boundary noise by embedding geometric regularity and integrating multi-scale, multi-resolution, and boundary-preserving mask features. AFMs play a crucial role in refining boundary precision by guiding feature extraction toward geometric consistency. As a result, our model achieves a 9.6% improvement in Average Precision (AP) and a 5% increase in Average Recall (AR) compared to the baseline, demonstrating its effectiveness in producing more accurate and regularized building footprints.
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.
How this classification was reachedexpand
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 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".