Why it matters
Without AI, misidentification of overpaint and restorations can lead to irreversible treatment damage, delayed loans, disputed insurance claims, and eroded client trust due to uncertain diagnostic evidence.
From design to safety
Design
AI-guided detection design
The system defines detection thresholds, reference libraries, and imaging protocols using prior restoration cases, pigment behaviors, and moveable retouching patterns to plan non-invasive assessment, with documented parameter sets, failure-mode considerations, and a requirement for pre-visit technical examination notes and target pigment pairs.
- UV fluorescence lamp
- X-ray cabinet
- Multispectral imaging system
Build
Equipment and protocol setup
Install imaging suite components, calibrate detectors, and configure data pipelines; formalise documentation for treatment proposals, condition reports, and PACR-aligned archival records, including version-controlled pipelines, audit trails, device registration, and staff training records, such as operator competency certificates.
- UV fluorescence lamp
- X-ray cabinet
- Multispectral imaging system
- Digital imaging workstation
Operation
Live detection and review
AI analyzes new scans, flags probable overpaint, fills, or retouching, cross-references historical notes, and generates a preliminary restoration history and treatment options; integrates with PACR-compliant condition logs and item-specific metadata schemas, and records review decisions.
- IR reflectography camera
- X-ray cabinet
- Multispectral imaging system
Maintenance
Calibration and data hygiene
Regularly recalibrate sensors, archive new reference samples, verify data integrity, and update detection thresholds with curator-approved changes; perform monthly backups and verify file integrity checksums, test disaster recovery, and refresh LUTs and spectral libraries.
- Digital imaging workstation
- X-ray cabinet
Safety
Safety and ethics governance
Ensure radiation safety, COSHH compliance, reversible detection practices, and adherence to PACR accreditation standards and professional ethics, including documented risk assessments and operator training, with annual safety drills and peer review of methodology.
- X-ray cabinet
- UV fluorescence lamp
- Multispectral imaging system
Part by part
- 1
UV fluorescence lamp
- Watch for
- Hidden retouching layers, underdrawing glow changes, fluoresced pigment boundaries indicating past retouching.
- AI action
- Flag suspected retouch zones; trigger review note and proposed reversibility assessment.
- 2
X-ray cabinet
- Watch for
- Metal-based pigments showing non-original structural supports; abnormal density indicating earlier interventions.
- AI action
- Highlight regions with unusual radiodensity for curator inspection and documentation.
- 3
IR reflectography camera
- Watch for
- Underdrawing and previous restorations revealed beneath pigmented layers.
- AI action
- Annotate underdrawn areas and compare with archival restoration history.
- 4
Multispectral imaging system
- Watch for
- Spectral anomalies suggesting different binders or overpaints across layers.
- AI action
- Generate composite maps showing layer relationships and potential restoration boundaries.
- 5
Digital imaging workstation
- Watch for
- Consistency in metadata, provenance fields, and versioned treatment proposals.
- AI action
- Produce a treatment-history log and export-ready condition report draft.
- 6
Treatment report binder
- Watch for
- Comprehensive archival of materials used and decisions made during restoration history.
- AI action
- Store final reports with traceable references and management of archive copies.
Worked example
Detecting a Hushed Retouch in a 17th-Century Portrait
Illustrative scenario with sample data.
A private studio applies portable X-ray and UV imaging to verify suspected overpaint from a mid-19th-century restoration.
Signal in
denomination: portrait; accessionNo: 1789-OP; uvFluorescence: strong green in background; xRayDensity: elevated in cheek region; historicalNotes: earlier repair 1850; canvas weave: 12-14 threads per cm; pigment set: lead-t white, vermilion, bone black.
System out
alert: probable overpaint detected in cheek region; recommended sampling plan; provisional restoration history added; next steps: confirm with cross-section analysis and archival comparison; report references archival notes 1789-OP-IR and 1850-Repair-Notes.
Human sign-off: Head Conservator reviews, signs off on treatment proposal, and updates condition report with note on reversibility and PACR-compliant archiving.
How to roll it out
Phase 1
Weeks 1 to 4: Baseline
Install imaging suite, load reference library, validate detection thresholds with a test suite, and document baseline condition for 20 selected paintings.
Phase 2
Weeks 5 to 12: Pilot
Run initial projects across two private collections and one public loan, refine reporting templates, and integrate with PACR-compliant archiving workflow and cross-reference schema.
Phase 3
Weeks 13 to 20: Full Operation
Scale to client projects, finalize maintenance schedules, validate long-term data integrity, and formalize safety and ethics assessments with annual audit.
Where people stay in charge
- Only a qualified conservator or conservation scientist may approve diagnostic conclusions and treatment proposals.
- All data handling, imaging, and radiation use must comply with COSHH, local regulations, and PACR accreditation requirements.
- AI-generated outputs must be reviewed against physical examination findings before any irreversible treatment decisions are made.
Questions owners and crews ask
What is the role of AI in detecting overpaint on master paintings?
AI supports the identification of non-original layers by correlating UV fluorescence, radiographs, and spectral data, but final decisions come from the conservator's assessment of material evidence.
What imaging modalities are involved for AI-assisted detection?
UV fluorescence, X-radiography, infrared reflectography, and multispectral imaging feed the AI system to map layer structure and retouching patterns for review.
How does AI influence treatment proposals and archives?
AI-generated detection maps and historical references guide treatment proposals and are stored in condition records and PACR-compliant archives for future reference.
Who signs off on AI-driven restoration history interpretations?
The Head Conservator or Conservation Scientist signs off, with curator and client input as needed, ensuring ethics and reversibility standards are met.