Deep learning-based 3D image reconstruction and damage mapping using neural radiance fields (Nerfacto)
Bibliographic record
Abstract
In structural health monitoring using computer vision, deep learning-based damage identification and three-dimensional (3D) reconstruction of the structure are current hot topics. Traditional photogrammetry techniques are cost-inefficient and time-consuming for 3D reconstruction, and there is no such solid 3D pixelwise damage mapping technique. To overcome these limitations, a new deep neural network (DNN)-based 3D reconstruction method, including damage mapping, is proposed in this article. As the DNN-based 3D reconstruction method, Nerfacto—an advanced version of Neural Radiance Fields models—was selected for achieving high-fidelity 3D reconstruction. This Nerfacto model was modified to create a high-definition 3D reconstruction model of the structure of interest (i.e., a 3-span bridge system). To map damages within the reconstructed 3D model using the modified Nerfacto, the state-of-the-art semantic transformer representation network (STRNet) with test time augmentation (TTA) was also developed for precise pixel-wise crack segmentation. Through extensive case studies, including parametric studies, we found that the modified Nerfacto can learn various appearance features of the structure and generate a very high-definition 3D model. Moreover, the segmented damage (i.e., cracks) from the STRNet with TTA could be mapped onto the reconstructed 3D model. This study demonstrates the potential of combining deep learning with 3D reconstruction for proactive and preventative maintenance strategies, ensuring the safety and longevity of vital structural assets.
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 | 0.001 | 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 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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 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".