Construction Industry Vision Alberta Dataset (CIVAD): Developing a Comprehensive Object Detection Dataset for Diverse Construction Applications
Bibliographic record
Abstract
Integrating computer vision technologies intothe construction industry has the potential to revolutionize site monitoring, safety management, and quality control.However, a critical gap remains in the availability of specialized datasets tailored to construction sites' distinct conditions and complexities while including sufficient classes representing most items in construction sites.Existing Computer Vision (CV) models often rely on generic training datasets, which limit their effectiveness for specific construction-monitoring-related tasks.Consequently, there is a pressing need for comprehensive domain-specific datasets that can capture the full spectrum of construction-related objects and activities.This study addresses this gap by developing a foundational training dataset called the Construction Industry Vision Alberta Dataset (CIVAD), specifically designed for CV applications in the construction sector.Our dataset included over 50 classes with more than 86,905 images of different objects, such as tools, machinery, safety equipment, and construction materials, to support diverse CV tasks.It utilizes a combination of web scraping, inclusion of existing open-source datasets, and direct data captured from construction sites.A set of novel methods, such as semiauto-labeling with advanced models, such as Grounded SAM and Grounding DINO, were used with our custom algorithms.These models were utilized to process parts of the dataset imagery with humans in the loop.This approach facilitated an efficient and accurate dataset creation process.The CIVAD dataset and methods employed in this study represent a significant step forward in integrating CV technologies across various construction-related applications.
Stored with the screening record, where it is evidence for the labels above.
How this classification was reachedexpand
The three-model screen
all 5,600 screened works →All three models called this out of scope.
Announcement of a computer vision training dataset for construction sites; a domain dataset, not research-system infrastructure.
The work develops a construction computer-vision dataset rather than studying research infrastructure or practice.
Construction-site computer-vision training dataset for industry monitoring, not research infrastructure as object.
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 | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.003 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.006 | 0.007 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.004 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.005 | 0.007 |
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".