Why it matters
Without AI, deterioration can advance unseen, leading to irreversible losses, delayed loans, disputed insurance, and erosion of client trust due to unclear condition histories and inconsistent monitoring.
From design to safety
Design
AI-supported treatment planning
Identify deterioration signatures, set acceptance criteria for imaging tests, formalize a data schema for condition reporting, PACR-aligned treatment proposals, and risk-based workflow mappings for loan readiness.
- high-resolution imaging station
- spectral camera
- documentation templates
Build
System acquisition and configuration
Select multi-sensor imaging (RGB, UV, IR, NIR), calibrate wavelength bands, install PACR-compliant condition reporting templates, and configure data routing to the digital archive.
- calibrated spectrophotometer
- imaging workflow server
- digital asset management
Operation
Active monitoring and classification
Process incoming scans, detect new defects, assign deterioration class, correlate with storage metadata, and generate preliminary notes for conservators; produce item-level deterioration summaries.
- microfilm scanner
- colorimeter
- object metadata ledger
Maintenance
Data hygiene and sensor upkeep
Run regular calibration, update deterioration models with reference samples, validate against standards, archive provenance with new condition readings, and refresh metadata schemas.
- lens cleaning station
- spectral calibration board
- quality control log
Safety
Object safety and ethics checks
Assess treatment risk, ensure non-destructive imaging remains within COSHH limits, document reversibility and minimal intervention considerations, and preserve professional integrity in reporting to clients.
- COSHH risk dossier
- benchtop laser safety gauge
- ethics review folder
Part by part
- 1
NIR-VIS Imaging Console
- Watch for
- Masking of subtle fading in 2D space, false positives in glare-heavy zones.
- AI action
- AI flags suspected dye fade or contrast loss; generates a jittered ROI for expert review.
- 2
Portable Microscope with Polarized Light
- Watch for
- Blooming on gelatin silver surfaces, micro-cracking in albumen layers.
- AI action
- Detects microstructural changes; prompts high-magnification review and chemical spot tests if needed.
- 3
Digital Asset Management Server
- Watch for
- Mislabelled scans, inconsistent metadata fields, missing accession numbers.
- AI action
- Centralizes metadata, enforces standardized fields, and links trends to storage conditions.
- 4
Portable Spectral Camera System
- Watch for
- Unusual spectral signatures suggesting dye fading or chemical alteration.
- AI action
- Routes flagged images to deterioration classifier and templates follow-up imaging protocols.
- 5
Consolidant and Cleaning Toolkit
- Watch for
- Inadequate vendor data on consolidant interaction with silver surfaces.
- AI action
- Traces material provenance and safety notes; flags potential incompatibilities for the treatment review.
- 6
Environmental Monitoring Suite
- Watch for
- Deviation from target RH/temperature in storage rooms.
- AI action
- Triggers alerts and archives environmental readings alongside deterioration observations.
Worked example
Detecting gelatin-silver image distortion before a loan
Illustrative scenario with sample data.
A private studio tracks fading in gelatin-silver prints and risks from fluctuating humidity. An AI-driven monitor flags subtle dye shift and spots potential silver mirroring early in the workflow before condition reporting.
Signal in
inputDate: 2026-08-22; accession: 2048-SEP; specimenType: gelatin-silver; storage: cold; RH: 35-40; scans: high-res, UV-vis; notes: prior micro-cracks noted.
System out
alert: deterioration-detected; defectType: dye-fade; location: image center; recommendedAction: confirm with high-res imaging and prepare treatment proposal; generatedReport: preliminary_condition_report_v1.pdf.
Human sign-off: Head of Conservation reviews the alert, approves deeper imaging, and authorizes a condition report update.
How to roll it out
Phase 1
Phase 1
Pilot the AI-assisted deterioration detection on a small batch of works on paper and photographs to validate classifier accuracy.
Phase 2
Phase 2
Scale to full cold-storage collection, establish baseline models, and integrate with condition reporting templates.
Phase 3
Phase 3
Formalize training, stakeholder approvals, and audit trails for PACR accreditation and client access.
Where people stay in charge
- Only a qualified conservator may approve treatment decisions based on AI outputs; final authority rests with the accredited conservator.
- AI outputs must be reviewed against archival metadata, material science data, and ethical considerations before any intervention.
- All risk assessments and reversibility considerations must be logged by the conservator before any action is taken.
Questions owners and crews ask
What data does the AI system read for photograph deterioration?
The AI reads spectral scans, high-resolution RGB images, metadata on storage conditions, accession details, and historical condition notes to identify deterioration signatures and trends.
How does the AI decide when to trigger a treatment proposal?
When the classifier detects a deterioration signal beyond predefined thresholds and correlates with material-specific risk factors, it flags the item and pre-populates a treatment-proposal template for reviewer sign-off.
Who can authorize actions suggested by the AI?
The accredited conservator signs off on treatment proposals and condition reports, with curator and client input as required by project scope.
What safety considerations are embedded in this workflow?
COSHH compliance, non-destructive imaging limits, reversibility criteria, and ethics reviews are required before any material interaction is approved.