Introduction of Deep Learning in Thermographic Monitoring of Cultural Heritage and Improvement by Automatic Thermogram Pre-Processing Algorithms
Bibliographic record
Abstract
The monitoring of heritage objects is necessary due to their continuous deterioration over time. Therefore, the joint use of the most up-to-date inspection techniques with the most innovative data processing algorithms plays an important role to apply the required prevention and conservation tasks in each case study. InfraRed Thermography (IRT) is one of the most used Non-Destructive Testing (NDT) techniques in the cultural heritage field due to its advantages in the analysis of delicate objects (i.e., undisturbed, non-contact and fast inspection of large surfaces) and its continuous evolution in both the acquisition and the processing of the data acquired. Despite the good qualitative and quantitative results obtained so far, the lack of automation in the IRT data interpretation predominates, with few automatic analyses that are limited to specific conditions and the technology of the thermographic camera. Deep Learning (DL) is a data processor with a versatile solution for highly automated analysis. Then, this paper introduces the latest state-of-the-art DL model for instance segmentation, Mask Region-Convolution Neural Network (Mask R-CNN), for the automatic detection and segmentation of the position and area of different surface and subsurface defects, respectively, in two different artistic objects belonging to the same family: Marquetry. For that, active IRT experiments are applied to each marquetry. The thermal image sequences acquired are used as input dataset in the Mask R-CNN learning process. Previously, two automatic thermal image pre-processing algorithms based on thermal fundamentals are applied to the acquired data in order to improve the contrast between defective and sound areas. Good detection and segmentation results are obtained regarding state-of-the-art IRT data processing algorithms, which experience difficulty in identifying the deepest defects in the tests. In addition, the performance of the Mask R-CNN is improved by the prior application of the proposed pre-processing algorithms.
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 | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| 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 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".