Why it matters
Without it, misinterpretation of damage, missed deterioration trends, or inconsistent annotations lead to delayed treatments, disputed insurance claims, and shaky client trust.
From design to safety
Design
AI informs design of documentation protocols
The system defines data schemas, damage-type ontology, and annotation standards from historical records and current imaging, guiding treatment proposal templates and archival metadata, ensuring provenance lineage and reversibility are embedded.
- High-Resolution Digital Camera Rig
- Annotation Tablet with Stylus
- Damage Type Ontology Database
- Change-over-Time Image Registry
Build
Workflow and hardware integration
Calibration routines read calibration targets, image captures align with metadata templates, and automatic tagging aligns with the damage ontology before it enters the archive, with project-structure templates.
- High-Resolution Digital Camera Rig
- Annotation Tablet with Stylus
- Damage Type Ontology Database
- Change-over-Time Image Registry
Operation
Real-time analysis of condition sets
AI analyzes new condition photographs, flags new or evolving damage, updates annotations, and generates interim condition reports for conservator review, plus a change-trend map for loans and exhibitions.
- High-Resolution Digital Camera Rig
- Annotation Tablet with Stylus
- Damage Type Ontology Database
- Change-over-Time Image Registry
Maintenance
Regular validation and data hygiene
Periodic recalibration of imaging, ontology updates, and integrity checks of change-over records preserve consistency across treatments and loans, with automated quarterly backup audits and versioning.
- Condition Photograph Archive Server
- Damage Type Ontology Database
- Change-over-Time Image Registry
- Conservation Treatment Report Generator
Safety
Safety, ethics, and accreditation guardrails
Qualified conservators review AI flags, ensure reversibility and minimal intervention principles, and archive decisions to satisfy PACR accreditation and COSHH compliance, including ethics reviews of automated annotations.
- Condition Photograph Archive Server
- Conservation Treatment Report Generator
- High-Resolution Digital Camera Rig
- Annotation Tablet with Stylus
Part by part
- 1
High-Resolution Digital Camera Rig
- Watch for
- miscalibrated color, lens distortion, inconsistent white balance, missing EXIF data
- AI action
- AI detects calibration drift and flags for re-shot or reprocessing, updating the asset metadata store
- 2
Annotation Tablet with Stylus
- Watch for
- missing annotations, ambiguous labels, inconsistent brush/pen IDs
- AI action
- AI suggests standardized labels and auto-embeds annotations into reports with version history
- 3
Damage Type Ontology Database
- Watch for
- new damage categories, ambiguous terms, conflicting definitions
- AI action
- AI maps observations to ontology terms and proposes updates for review, maintaining a revision log
- 4
Change-over-Time Image Registry
- Watch for
- non-sequential time stamps, gaps in imaging series, divergent file formats
- AI action
- AI sequences images by date and prompts checks for continuity, alerts missing entries
- 5
Condition Photograph Archive Server
- Watch for
- data corruption, access control violations, backups lagging
- AI action
- AI flags integrity issues and initiates backup/restore workflows, records access events
- 6
Conservation Treatment Report Generator
- Watch for
- inconsistent terminology, missing method details, non-reproducible steps
- AI action
- AI assembles treatment reports from annotated images and ontology terms, preps draft for sign-off
Worked example
Catching a Surface Abrasion Pattern Drift
Illustrative scenario with sample data.
A private studio tracks abrasion progression on a wooden panel across five years, using AI to compare condition photographs and annotate changes.
Signal in
field: captureDate=2026-03-12; cameraModel=DSLR-Pro; lighting=5500K; exposure=1/125s; ISO=400; magnification=1.5x; datasetID=IMG-20260312-PA-01; previousID=IMG-20220112-PA-01
System out
alert: Abrasion area drift detected in Panel A between 2021 and 2026; annotated change map created; suggested inspection interval 6 months; draft treatment proposal generated for review.
Human sign-off: Head of Conservation reviews the AI-generated map and annotations, approves the draft treatment proposal, and signs off the final condition report.
How to roll it out
Phase 1
Weeks 1 to 4
Baseline imaging, ontology alignment, and annotation template definitions established.
Phase 2
Weeks 5 to 8
Workflow integration, calibration routines, and issue-tracking dashboards activated.
Phase 3
Weeks 9 to 12
Pilot projects commence with select condition sets and report generation validation.
Where people stay in charge
- All AI-derived decisions must be reviewed and signed by a qualified conservator before archival entry.
- AI-generated annotations and ontologies require human validation for reversibility and ethics compliance.
- The system will not modify treatment proposals or access controls without the accredited conservator's approval.
Questions owners and crews ask
What counts as a condition photograph in this system?
Condition photographs are standardized images captured under controlled lighting with fixed angles, calibrated color targets, and complete metadata linked to the object's condition report.
How does AI handle new damage types?
AI maps observations to an existing ontology and prompts human review; if novel types recur, ontology terms are updated after expert validation.
Who signs off on AI-generated reports?
The accredited conservator or head of conservation signs off; clients receive the finalized condition reports with AI provenance notes.
How is data privacy protected for loaned objects?
Access controls, audit trails, and encrypted storage ensure only authorized staff can view condition records and imaging datasets.