Road Network Extraction Using CNN Architectures
Bibliographic record
Abstract
The rapid advancement of satellite imagery and deep learning technologies has opened new avenues for automated geospatial data extraction and mapping. This research presents a novel approach to road network extraction from high-resolution satellite imagery, leveraging deep learning techniques and the Spacenet dataset to achieve over 99% of accuracy and DICE coefficient without overfitting by accurately identifying and delineating road networks from satellite imagery, and moreover, presenting an approach to optimize the post-processing with the ultimate goal of contributing to open-source mapping platforms like OpenStreetMap (OSM). Also, the manuscript highlights challenges and opens avenues for future research, including developing new metrics akin to the DICE coefficient and optimizing computational efficiency for model training with limited resources. By implementing and comparing three distinct Convolutional Neural Network (CNN) architectures using deep learning capabilities, the research systematically evaluated the performance of road network extraction techniques. The proposed methodology encompassed comprehensive data pre-processing, advanced deep learning model training, and post-processing strategies to transform raster road network predictions into vector data suitable for geospatial analysis. The research workflow demonstrated a seamless integration of satellite imagery analysis, deep learning road extraction, and open-source mapping platforms, thereby advancing automated geographic information system (GIS) methodologies with vectorized outputs suitable for seamless integration into mapping software platforms. The manuscript highlights the potential of integrating deep learning techniques with professional GIS software to enhance road network mapping and contributes to the expanding field of automated geospatial intelligence.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.002 |
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 source (direct Gemma or distilled Codex), 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".