Why it matters
Without AI, uncertain dirt composition and bloom can lead to overcleaning, solvent mishaps, or missed signs of varnish degradation, creating irreversible risk and lost client trust.
From design to safety
Design
AI-driven condition mapping for proposals
The system reads historical condition data, spectra, and imaging studies to propose cleaning plans, safety margins, and material choices with traceable justifications. It expands its model with reference libraries, reversibility goals, and explicit ethical constraints, incorporating conservator feedback loops and long‑term archival considerations.
- Raking Light Imaging System
- UV Fluorescence Lamp
- Infrared Reflectography Camera
Build
Integrating imaging pipelines and decision rules
Establish data schemas, calibrations, and workflow rules that map surface observations to cleaning strategies, test logs, and documentation formats. It adds provenance tracking for every decision and on-site QA checks, plus hardware integration with environmental sensors.
- Digital Microscope
- Cleaning Solvent Cabinet
- Microfade Tester
Operation
Real-time dirt and bloom classification at bench
The AI processes current images, colorimetry, and spectroscopic cues to categorize dirt types and bloom presence, suggesting tests and safe solvents. It produces management-ready reports for condition status and treatment planning, with versioned treatment proposals.
- Raking Light Imaging System
- UV Fluorescence Lamp
- Digital Microscope
Maintenance
Data hygiene and calibration audits
Regularly recalibrate imaging devices, refresh training data, and verify AI decision thresholds against new reference samples and test logs. It schedules quarterly audits and stores versioned rule sets, with change logs and backup plans.
- Infrared Reflectography Camera
- Microfade Tester
- Cleaning Solvent Cabinet
Safety
Preserving conservation ethics and safety
The system flags irreversible treatments, ensures reversibility, and logs COSHH considerations, with final sign-off by the accredited conservator. It aligns with PACR standards for documentation, staff training, and ethical review workflows.
- UV Fluorescence Lamp
- Raking Light Imaging System
- Digital Microscope
Part by part
- 1
Raking Light Imaging System
- Watch for
- New or existing surface textures changing in raking light, indicating dirt or varnish anomalies.
- AI action
- AI detects texture shifts and flags likely dirt vs. varnish bloom for inspection and targeted sampling.
- 2
UV Fluorescence Lamp
- Watch for
- Fluorescent signatures of varnish bloom or wax/resin residues indicating contamination.
- AI action
- AI maps fluorescence patterns to probable bloom zones and triggers targeted tests and solvent compatibility checks.
- 3
Infrared Reflectography Camera
- Watch for
- Subsurface residues or previous restorations affecting cleaning strategy.
- AI action
- AI highlights subsurface contrasts to guide safer cleaning paths and avoid over-cleaning.
- 4
Digital Microscope
- Watch for
- Micro-areas of surface dirt, micro-cracks, or PVA/resin deposits.
- AI action
- AI quantifies particle size and distribution to tailor testing sequence and record microscopic evidence.
- 5
Cleaning Solvent Cabinet
- Watch for
- Solvent compatibility alerts with surface layers and varnish compositions.
- AI action
- AI recommends solvent sequence and safety margins before application, logs solvent exposure times.
- 6
Microfade Tester
- Watch for
- Smaller color changes indicating pigments vulnerable to solvent exposure.
- AI action
- AI forecasts color change thresholds to constrain cleaning tests and document results.
Worked example
Assessing dirt and bloom on a 19th-century scene
Illustrative scenario with sample data.
A private studio evaluates a 19th-century painting with surface dust and yellowing bloom affecting glaze layers.
Signal in
field:data|image:diagnostic|condition:oxidized varnish|spectra:IR-SS|testLogs:pending|treatmentProposal:required
System out
alert: bloom localized in sky area; action: test log entry created; recommended cleaning sequence entered into report; future QA check scheduled
Human sign-off: Head conservator approves cleaning test sequence and validates the treatment proposal.
How to roll it out
Phase 1
Phase 1
Baseline implementation for imaging, logging, and basic AI classification in studio workflows.
Phase 2
Phase 2
Expanded rule set, calibration against reference libraries, and integration with condition reports.
Phase 3
Phase 3
Full QA checks, on-site testing, and archival of AI decisions within PACR-compliant records.
Where people stay in charge
- All cleaning decisions involving chemical interventions must be reviewed and signed off by a qualified conservator.
- AI classifications are advisory; final treatment choices require human judgment and ethical review.
- All data handling and imaging must comply with privacy, safety standards, and institutional policies.
Questions owners and crews ask
What kinds of dirt can AI identify on paintings surfaces?
The system recognizes dust, soot, oily residues, wax or resin bloom, and micro-particles, distinguishing them from varnish contamination or pigment alteration through imaging cues.
How does AI influence treatment proposals for paintings?
AI provides evidence-backed classifications and recommended test orders, which inform conservative, reversible treatment proposals with traceable justifications.
What records are produced for condition documentation?
Treatment proposals, test logs, digital imaging records, and AI decision notes become part of the condition reports and conservation archives.
Who signs off on AI-generated decisions?
Accredited conservators review and sign off on AI-assisted treatment plans and safety assessments before work proceeds.