A Deep Learning-based Surrogate for the XRF Approximation of Elemental Composition within Archaeological Artefacts before Restoration
Bibliographic record
Abstract
The restoration of archaeological artefacts is naturally utterly important for preserving the cultural heritage. The first step that is undertaken in this process is the chemical analysis of the object, in order to decide the best procedures for its restoration. The gold standard in approximating the concentration of the elements in its composition (in percentages, between 0 and 100) is performed through an X-ray fluorescence (XRF) machine. While this is a non-invasive approach, it comes at substantial financial and training costs, and possible radiation exposure of the investigator. In this context, the present paper explores the potential of a deep learning regression model to give an estimate on the concentration of a given element from stereo microscopy slides of historical artefacts, as an alternative means to the XRF. Two problems with different degrees of complexity are examined in turn. The first one is represented by the consideration of iron objects, where the metal is strongly dominant in the chemical structure. The second comes both as a complement to the other, in order to expose the model also to non-iron items, and as a more difficult task of identifying the degree of copper that is present only as part of an alloy constitution. While for iron the one absolute value prediction of the model is always very close to the XRF approximation, copper has a wider distribution of its concentration among objects, which is more challenging to learn; hence, performance for a singular absolute estimation can rise only with the increase in the amount of data. A window of error acceptability was also implemented and it allows for an approximation that is sufficient for grasping the degree of the metal in the composition that is necessary for the restoration procedures. The findings therefore provide a first step in putting forward a computational support tool that represents a less expensive and less dangerous alternative for approximating the elemental analysis before artefact reinstatement.
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 | 0.003 |
| 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 | 0.001 |
| 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 | 0.001 |
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".