Deep learning-based rebar detection and instance segmentation in images
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
• A high-quality image benchmark is developed for rebar detection and segmentation. • The dataset covers two types of rebars under various shooting views and placement patterns. • The comparison test provides the basis for selecting deep learning models. • Six data augmentation strategies are discussed to improve the best models. • Three downstream tasks in real-world scenarios validate the developed model. Automated rebar cage assembly and quality inspection rely on reliable rebar perception. Recent studies have explored image-based rebar perception via object detection and instance segmentation algorithms. However, existing models are limited across various scenarios, especially with different rebar categories, arrangement patterns, and camera views, which limits their application. This is primarily attributed to the absence of a benchmark considering these factors. This study introduces an image benchmark designed for the effective training and selection of rebar detection and instance segmentation algorithms. It is the first to encompass two types of commonly used rebars, multiple camera views, and various rebar placement patterns at different assembly phases in a single dataset. Six object detection methods and four instance segmentation methods are evaluated to assess the applicability of the state-of-the-art methods. Additionally, a new shape-prior-based post-processing method is developed to address the merged detection problem in clustering. The experiment shows that Deformable DETR and Mask2Former achieved the highest bounding box mAP (80.4) and mask mAP (66.3) respectively. The Simple Copy-Paste technique was introduced, improving the mask mAP of Mask2Former by 2.8 points. Finally, the developed model was validated in the real-world scenarios of three downstream tasks. Notably, in the rebar spacing measurement task, the proposed post-processing method improves Mask2Former by increasing its bounding box mAP by 18.0 and mask mAP by 2.4.
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 | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.000 | 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