Interpretation Status

This page tracks how far OpenMeta has moved beyond raw metadata decoding into meaningful interpretation. Interpretation means that decoded entries have stable names, typed values, semantic groups, query shapes, and transfer-safety classification that host applications can use directly.

Current overall status: high, measured above 99% for the public target scope. This is intentionally lower than decode coverage. Decode parity only proves that metadata carriers and entries are visible; interpretation also requires human-readable meaning and safe cross-format behavior.

100% acceptance gates

For the declared target scope, an entry counts as covered when it has one explicit outcome:

Gate	Requirement
Decoded	The carrier and entry are visible in MetaStore, or the decoder reports an explicit unsupported/limit/malformed reason.
Named	The entry has a stable public name, or it is deliberately exposed as an unknown numeric/private field.
Typed	The raw value shape is preserved as scalar, vector, matrix, table, bytes, text, or opaque blob.
Interpreted	Known enum-like values, orientation states, geometry, exposure/gain, color, white-balance, lens-correction, RAW-processing, and source-private meanings are projected into public helpers or query candidates.
Classified	Source-bound data is classified as portable, target-owned, source RAW-specific, vendor-private, computational, thermal, preview/face/ stitch metadata, or opaque/lossless.
Queryable	Host/UI workflows can find the interpreted meaning through focused query helpers with source entries, confidence, value shape, and normalized fields where available.
Fuzzy-searchable	Optional search can tolerate bounded spelling and property-path near misses while reporting similarity and exact/fuzzy provenance without silently changing interpretation.
Structured	Host code can consume query-backed interpretation records without reassembling raw query candidates manually.
Conflict-aware	Duplicated cross-family concepts either have a documented precedence rule or surface enough source information for host conflict handling.

Measured target audit

The current target audit measures 99.85% interpretation coverage and 99.77% query coverage. The denominator contains only explicitly declared semantic targets. Unknown private numeric IDs and intentionally opaque payloads are reported separately, while malformed, undefined, and incomplete values remain visible as residual gaps instead of receiving guessed meanings. Fuzzy Search is measured separately and is not part of the query percentage.

Coverage matrix

Area	Current coverage	Readiness	Main remaining gap
Standard EXIF/TIFF/DNG tag names and typed values	Standard tag names, common scalar/vector values, DNG crop/color/ exposure/RAW-processing fields, GeoTIFF key names, EXIF 3.1 learning/ development/correction/noise tag names, and common EXIF/TIFF/DNG numeric value-name helpers are available. Exposure time, aperture, ISO sensitivity, exposure bias, exposure program/mode, gain, correction/noise status values, and raw exposure-adjustment records now flow into concept candidates or stable display helpers where appropriate.	High, above 99% for declared enum targets.	Undefined enum codes and malformed field encodings intentionally remain raw.
ICC profiles	ICC header/tag table decode plus interpreted desc, text, signatures, XYZ, curves, named-color, measurement, viewing-condition, MFT/MAB/MBA, numeric array, and malformed/limit handling.	High, about 90-95%.	Full color-management policy remains host-owned; OpenMeta interprets profile metadata, not rendered color transforms.
IPTC-IIM and portable XMP	IPTC datasets and XMP properties decode into typed entries, bounded EXIF/IPTC-to-XMP projection exists for transfer/writeback, and common descriptive EXIF/IPTC/XMP concepts such as title/headline, description/caption, creator/author, keywords/subject, rights/license, credit/source, and IPTC date/time promoted into cross-family created-date candidates are queryable with source-entry provenance. The Descriptive concept also reconciles equivalent legacy IPTC/Photoshop fields; preserves non-equivalent object-type, object-attribute, and subject references; scopes editorial status/update/action/cycle/language/fixture fields, content locations, prior-envelope references, originating software, and editorial contacts; and normalizes editorial release/expiration date-time pairs. Legacy image type/layout, rasterized caption, audio type/channel/content/rate/resolution/duration/outcue, and object-preview format/version/data datasets are interpreted as source-bound technical records; IPTC image layout is not conflated with EXIF rotation orientation. It also interprets IPTC Core accessibility/taxonomy fields, IPTC Extension controlled-vocabulary, registry, image-region entity records, and rectangle/circle/polygon boundary geometry with explicit pixel/relative units, XMP document identity, and XMP Media Management resource-reference, ingredient, history, manifest-item/reference, version/event, and bounded pantry records; and groups structured creator contacts, events, people, organizations, products, artwork/objects, encoded rights expressions, PLUS parties/assets/ license policy, and model/property releases with record kind, record scope, language, transfer hint, and independent sensitivity. XMP CreatorTool is interpreted as source software, Camera Raw Settings requires the exact standard namespace, and unknown nested namespaces retain full URI identity in collision-safe nsu_<namespace-uri-hex>: path segments.	High, about 99%.	Deliberately unmodeled arbitrary pantry payloads and undocumented extension datasets remain bounded raw metadata.
Orientation	EXIF/TIFF orientation query, LibRaw flip mapping, and generic orientation helpers for index, rotation degrees, mirrored state, dimension swap, rotation-only fallback, human-readable labels, and EXIF-vs-XMP conflict reporting in the LibRaw bridge.	High, about 90-95%.	Higher-level policy for resolving container and host pixel-orientation state remains host-specific.
Geometry, crop, active area, and borders	DNG crop/active-area/masked-area tags, Phase One/Leaf geometry, Fujifilm RAF raw crop/zoom rectangles, Canon aspect/crop metadata, Nikon Capture crop bounds, Sony panorama crop margins, canonical border margins, vendor RAW-processing geometry buckets, and crop/border-style paths are queryable.	High, about 88-92%.	More vendor-specific normalized rectangles and stronger output contracts for ambiguous multi-tag geometry.
Exposure and gain	Standard EXIF exposure time, f-number, exposure program/mode, photographic sensitivity, exposure bias, exposure index, gain control, selected DNG baseline/raw-preview gain fields, matching XMP paths, and selected decoded vendor/MakerNote exposure names are queryable and promoted into cross-family exposure roles. Standard EXIF exposure program/mode and gain-control values plus selected Canon/Nikon/Sony/Fujifilm/ Pentax/Olympus/Panasonic/Phase One/Kodak/Minolta/Sigma/Samsung/Ricoh MakerNote values carry human-readable labels where stable. Capture exposure facts are marked safe, while raw/DNG exposure adjustments are marked unsafe for rendered-image transfer.	Medium-high, about 91-94%.	More vendor MakerNote exposure print conversions and richer per-vendor exposure/gain labels.
Color, white balance, profiles, and matrices	DNG color/calibration/reduction/forward matrix groups, white-balance vector groups, EXIF color-space evidence, ICC header/tag entries, XMP ICC/profile fields, PNG profile text carriers, RAW color/source- processing safety buckets, transfer hints, per-family grouped vendor color/WB candidates, long-tail camera-to-XYZ/RGB, Canon ColorData source-color tables, style/color, and white-balance gain aliases, and cross-family concept candidates with full grouped value vectors are identified. ICC/profile and color-space records have a distinct color_profile semantic role, while camera RAW profile/look/ tone-curve/style fields and vendor source color tables have a separate source_color_transform role marked unsafe for rendered-image transfer. Matrix/vector groups require numeric payloads with conservative minimum shapes before promotion.	Medium-high, about 86-92%.	Deeper camera/vendor color science interpretation is intentionally conservative, especially for rendered-image transfer.
Lens correction and RAW processing	Lens-correction groups, black/white levels, linearization, RAW value curves, RAW linearity limits, RAW calibration curves, RAW curve control points, CFA/sensor layout, raw-storage identifiers, vendor RAW/source- processing buckets, creative/picture style, film simulation, dynamic-range, optical correction, raw-development, computational, thermal, and stitch/panorama aliases, per-family vendor raw-storage/ sensor/computational/thermal/stitch/source-processing table candidates, transfer hints, transfer diagnostics, and concept candidates with grouped table/vector values are classified for query and transfer safety. Current public RAW curve coverage includes DNG linearization/ linearity-limit tags plus conservative Sony, Nikon, Kodak, Panasonic, and Phase One/Leaf-style curve or calibration names where decoded metadata is visible. Sony model-specific correction-offset routing includes the current ILCE-7RM6 Tag9416 offsets where the decoded payload exposes numeric correction arrays. RAW concept candidates now expose a conservative RAW applicability state; descriptor-aware concept-resolution overloads let a host or decoder provide a MetadataRawDataDescriptor so curve/LUT-like entries can remain conditional, apply to stored RAW samples, require compressed RAW storage, require the primary RAW plane, or be marked not applicable for rendered, uncompressed, packed, or non-primary-plane data. Transfer preparation can also consume the descriptor and drop RAW-processing metadata when source pixels are already rendered. Lens-correction grouped tables require numeric payloads before promotion.	Medium-high, about 90-94%.	Long-tail per-model correction tables, exact blob/decoder-stage applicability, and richer numeric normalization.
Vendor MakerNotes	Broad MakerNote naming and source-processing classification exists for common vendors and several live computational/thermal vendors. Unknown entries remain lossless and source-private subgroups distinguish preview, face geometry, computational, thermal, stitch/panorama, pixel-shift, multi-shot, composite, auto-lighting, RAW crop/active-area, source color-transform, source style/rendering aliases, lens-correction, raw-level processing data, and Phase One/Leaf RAW-processing fields handled by direct classification plus dedicated normalized helpers. Classified multi-field vendor groups now surface as grouped query/interpretation candidates where safe to expose structurally. Selected Canon/Nikon/Sony/Fujifilm/Pentax/Olympus/Panasonic/Casio/ Phase One/Kodak/Minolta/Sigma/Samsung/Ricoh/Apple/FLIR/JVC/General Imaging/Reconyx/Microsoft/Motorola/Nintendo/Sanyo print conversions expose bounded human-readable labels, including expanded Canon sub-IFDs and CanonCustom fields, Canon ColorData source color-transform aliases, Nikon sub-IFDs and NikonSettings fields, NikonSettings source-processing aliases, NikonSettings On/Off residual labels, Nikon Active D-Lighting labels, decoded Fujifilm mk_fuji* including flash white-balance naming, native RAF firmware naming, Pentax sub-IFDs, Olympus main/focus/equipment fields, Casio Type2 fields, Panasonic long-tail main-table fields, Apple AE/AF/HDR/capture/camera-type fields, FLIR GPS-valid state, JVC quality, General Imaging macro state, Reconyx moon phase/weekday/flash/illumination/battery/trigger labels, Microsoft stitch camera-motion/map-type labels, Motorola CustomRendered labels, Nintendo category labels, Sanyo main/MOV public-context scalar labels, current Canon RF lens-type labels, current Nikon Z LensData0800 LensID labels, and an ambiguous Pentax Sigma/Samsung/Tokina lens-family label. Version/firmware payloads have a separate bounded formatter path for selected standard EXIF byte-version fields, Nikon version-like contexts, Olympus packed firmware fields, and native RAF firmware fields so formatted versions are not confused with enum labels. Canon AF micro-adjustment fields are classified for lens-correction search, and Canon ambience-selection fields are classified as source-processing metadata. Ambiguous per-model, per-version, text/count, measurement, MP4 numeric, or value-type-dependent labels intentionally remain empty instead of guessing.	Medium-high, about 97%.	Remaining encrypted/custom settings, per-model private tables, remaining live-vendor scalar/string-coded fields, and per-model firmware formulas outside currently supported contexts.
BMFF item graph, HEIF/AVIF/CR3, JUMBF, and C2PA	BMFF derived fields, brand-name fields including avif / avis / avio AVIF-compatible brands, item-info rows, item type/semantic labels and semantic aggregate counters for common metadata carriers, whole-scene item graph counts, bounded graph component summaries, bounded per-component rows with role text, ordered member item IDs, known/unknown and semantic node counts, isolated-state, primary membership, typed auxiliary/derived/thumbnail/ content-description relation counts with direction-correct endpoint roles and named item-id aliases, alpha/depth/disparity/matte auxiliary counts, and conservative component policy text, component content-bound counts, primary component content-bound flags, primary component multi-image candidates and policy text, and conservative content-bound metadata / multi-image policy hints, primary metadata-carrier/C2PA/JUMBF flags when the primary item itself is a metadata item, primary sidecar counts/flags for linked metadata and image sidecars, separate primary inbound derived-item and outbound derived-source summaries, content-bound C2PA/JUMBF sidecar policy hints, compact primary-scene node/edge summaries with unique linked-item role buckets, per-role edge counters, and image/metadata/content-bound metadata node counts, bounded ipco property-container summary counts, bounded ipma item-property association rows and per-property-type association/primary/essential rollups, bounded relations, grpl item-group rows with group semantics and ordered entity roles for altr, ster, and pymd, per-group-type summaries, primary item-group memberships, bounded iloc/idat item-data layout summaries, bounded grid/iovl/iden construction descriptors with ordered source IDs, grid coordinates, overlay offsets/background, identity sources, recursive method-2 item-offset descriptor reads, descriptor-reference depth, and fail-closed graph cycle/missing-source/ truncated-reference validity fields. tili image items expose bounded tilC version-0 dimensions, extra dimensions, property relationships, tile grids, expected tile counts, dref/deti mapping, internal tile_item_type/tipa associations, external URL state, logical offset tables, explicit or sequentially inferred tile sizes, empty-tile state, separate core/layout/complete validity, and a source-bound container-graph concept. Primary item-location and derived-construction aliases, primary-linked roles with linked-item semantic aggregate counters, aux semantics, primary color/profile property summaries, primary display dimensions and transform summary, primary pixel aspect ratio, primary pixel component bit depth, clean-aperture rationals, JUMBF box labels, and draft C2PA/JUMBF structural fields are exposed.	High, about 98%.	Independently authored tiled-image conformance files and full C2PA manifest/policy semantics.
Photoshop IRB	Raw resources are preserved and a bounded interpreted subset is decoded for fixed-layout resources, including Photoshop 2 info/color-table summaries, resolution/version/print data, print-flag bytes, border/background/effective-BW data, display info, grid/guide info, color sampler headers/records, descriptor-header summaries plus safe descriptor class-name/class-ID/item-count fields, bounded descriptor item bodies (bool, long, comp, doub, UntF, TEXT, enum, type, GlbC, opaque alis and tdta byte counts) with type-name/type-code fields, parsed maximum depth, and parsed per-type counters, ordered bounded obj reference streams for property, class, enumerated, offset, identifier, index, and name forms with reference paths, depths, indices, values, and aggregate counters, plus empty and non-empty nested object/list summaries with item paths, depths, list indices, and parsed-value counts, for resources including layer comps, measurement scale, timeline info, sheet disclosure, HDR toning, print info, onion skins, count info, print info/style, path selection state, and origin path info, working-path and numbered clipping-path byte-count / record summaries, alpha names/identifiers, captions, QuickMask info, URL/list data, autosave strings, XMLData, ImageReady XML text, Lightroom workflow text, thumbnail headers, channel options, clipping-path names, Macintosh PrintInfo, Macintosh NSPrintInfo, Windows DEVMODE, alternate duotone-color, alternate spot-color, and obsolete Photoshop tag byte counts, legacy halftone/transfer/duotone/EPS byte summaries, embedded IPTC/ICC/XMP/EXIF resource byte counts, and embedded IPTC/XMP/ICC payload decode where enabled.	Medium-high, about 90-94%.	Full Photoshop action execution semantics, opaque alias/raw payload interpretation, and long-tail resource interpretation.
Semantic query and records	Query helpers expose raw matches, confidence, provenance, value shapes, normalized candidates, canonical crop/active-area rectangles, Fujifilm RAF raw crop/zoom rectangles, Canon/Nikon/Sony crop and border patterns, border margins, exposure/gain roles, selected vendor/MakerNote exposure-name aliases, per-family grouped vendor records, descriptive EXIF/IPTC/XMP concepts including exact rights/license/credit/source/editorial/accessibility/taxonomy/registry/ image-region/document-identity/document-lineage/document-history/ technical-image/audio/preview semantics, explicit color-profile records for EXIF/ICC/XMP/PNG profile carriers, explicit source-color-transform records for camera RAW profiles, looks, tone curves, Canon ColorData tables, and vendor source color tables, explicit raw_value_curve, raw_linearity_limit, raw_calibration_curve, and raw_curve_control_points records, explicit computational/thermal/stitch/source-processing records including NikonSettings groups, expanded source color/style/lens/ source-processing aliases including Canon AF micro-adjustment and ambience-selection fields, source-processing buckets, explicit container_graph concepts for BMFF content-bound metadata, derived-image construction, tiled-image configuration, and whole-scene/primary-component/ per-component multi-image policy, structured interpretation records, and bounded cross-family concept resolution for orientation, date/time, exposure/gain, color/profile, GPS, descriptive fields, geometry, lens-correction, RAW-processing, and container graph policy with parsed date/time fields, matching EXIF OffsetTime*/ SubSecTime* provenance, bounded subsecond precision, normalized-instant conflict checks, IPTC created and digital-creation date/time plus XMP DateTimeDigitized promoted into cross-family date candidates, same-scope EXIF/XMP GPS timestamp assembly, distinct camera/destination/shown/created coordinate roles, scope-aware structured-location conflict handling, language-aware descriptive scalar conflicts, structured descriptive record kinds/scopes, independent sensitivity, GPS altitude-reference state and display token, canonical geometry origin/size/rect/margins, normalized exposure values, shape-checked grouped value vectors, transfer hints, RAW applicability states, rendered/compatible safety booleans, and tolerance-aware GPS/exposure/color/geometry conflicts.	High, measured about 99.77% for declared targets.	Incomplete GPS tuples, malformed dates, and undefined values remain inspection data rather than normalized candidates.
Fuzzy Search	Optional fuzzy_search_metadata(...) searches decoded names and property paths with bounded score/result options, deterministic top-k ordering, stable ties, exact/alias/fuzzy provenance, curated positive/adversarial quality gates, and an opt-in scaling benchmark. Python snapshot/document wrappers preserve the same result shape.	High enough for the current milestone, about 80-85%.	Broaden independently sourced quality coverage, then design Unicode normalization, transliteration, multilingual ranking, and an optional immutable index for repeated large-store searches.
Transfer-safety classification	Compatible-file versus rendered-image safety policies classify source-specific image geometry, color/profile, RAW curves/linearity metadata, RAW-processing, MakerNote, BMFF content-bound/multi-image/ derived-construction/tiled-image policy, JUMBF/C2PA, and vendor-private data, with concept-level diagnostics that report keep/drop/requires-target-image-spec actions, severity, stable summary and message tokens, localizable argument tokens, RAW applicability, and role-specific default message text before prepare. BMFF content-bound/multi-image scene, derived-image construction, and tiled-image configuration fields resolve to source-bound container-graph diagnostics for rendered-image transfers. Source-processing diagnostics distinguish computational, thermal, and stitch/panorama message tokens. Descriptor-aware diagnostics can also mark curve/LUT-like RAW roles as compressed-storage-only or primary-plane-only. PrepareTransferRequest can carry a MetadataRawDataDescriptor; when it marks source pixels as rendered, RAW-processing metadata is filtered even under compatible-file safety.	High, about 94-96%.	More per-family policy tests and broader per-family policy coverage.

Competitor position

ExifTool remains the practical reference for long-tail tag names, MakerNote tables, and human-readable print conversions. OpenMeta is now close on decode visibility for the current target scope, but interpretation still trails ExifTool in per-model private meanings.

Exiv2 is strong for common EXIF/IPTC/XMP workflows. OpenMeta’s differentiator is the explicit safe-transfer and host-query model: it classifies whether data is portable, target-owned, source RAW-specific, or unsafe to move into rendered outputs.

Active project sequence

The first bounded Creation and Editing milestones are complete. The current implementation focus advances to Transfer, Translation, and Writing. Adapters and Utilities remain deferred. Creation and Editing resume for custom, multilingual, structured, direct-family output, structural block operations, and broader synchronization; Fuzzy Search resumes before those final two stages for its Unicode/multilingual and optional-index milestones.

Interpretation maintenance priorities

1. Validate the complete bounded BMFF tiled-image contract against independently authored conformance files as they become available.

2. Broaden transfer diagnostic policy coverage now that stable message tokens and localizable argument tokens are available for GUI workflows.

3. Add pantry payload semantics only where a validated, bounded schema exists; keep arbitrary payloads raw and source-bound.

4. Expand the remaining unambiguous MakerNote long tail: encrypted/custom settings, per-model firmware formulas outside currently supported formatter contexts, remaining live-vendor scalar/string-coded fields, and per-model tables only where context is strong enough to avoid wrong labels.

5. Keep transfer-safety classification conservative when interpretation is incomplete.