new → triaged → uploaded → enriched → extracted → ▶ closed · remove · ignored
closed
ARK
QR code for image 01KQJA2XDVE9WAR99MFQTG7F9Z
Name
2DbdH5AbBwoEv33bunKxH_
https://n2t.net/ark:/11635/2DbdH5AbBwoEv33bunKxH_
https://scanstation.ai/ark/11635/2DbdH5AbBwoEv33bunKxH_
File
Order 46
Dimensions 7968 × 5984
Orientation 0
S3 Key marac/003/derivatives/0046.jpg
Source Path marac/003/derivatives/0046.jpg
AI Enrichment
Summary
Thumbnail shows the back of a divided postcard mounted in a ScanStation capture setup, with a red postage stamp, possible postmark, handwritten message on the left, and a handwritten address on the right; the text and postal details require high-resolution review for reliable transcription.
Places
postal destination on address side
Type
tag_or_label document text_page postcard
Tags
No tags assigned.
Primary: postcard
No description yet.
Subject: image:01KQJA2XDVE9WAR99MFQTG7F9Z 17 predicates
Predicate Best value Conf. Source
ai:denseSummary
Thumbnail shows the back of a divided postcard mounted in a ScanStation capture setup, with a red postage stamp, possible postmark, handwritten message on the left, and a handwritten address on the right; the text and postal details require high-resolution review for reliable transcription.
1% [email protected]
ai:documentLikely yes
ai-tools heuristic on scan.storage:marac/003/0046.jpg
1% [email protected]
ai:handwrittenText yes 1% [email protected]
ai:hasText yes 1% [email protected]
ai:highResGoal
read_handwriting
Loss of sender/recipient names, place names, message content, postal markings, and possible dating evidence visible only in the handwriting and stamp/postmark.
1% [email protected]
ai:highResTarget
handwritten message on left half of postcard
High-resolution target from thumbnail enrich
1% [email protected]
ai:isFilledForm
ai:isForm
ai:ocrText
Tesseract via ai-tools (6 words, mean conf 34.3)
0% [email protected]
ai:peopleLikely
ai:photoLikely
ai:photoWithTextLikely
ai:pixelsDone
ai:speculation
This is likely a documentary digitization image rather than a consumer scan, made with a repro station using registration targets.
Visible fiducial targets, orientation arrow, and the 'ScanStation' label indicate controlled capture equipment.
1% [email protected]
ai:typedText yes 1% [email protected]
dcterms:spatial
postal destination on address side
1% aggregated
dcterms:type
tag_or_label
found 3 machine-readable code(s); small text footprint matches labels or tags
5% [email protected]
Subject: image:01KQJA2XDVE9WAR99MFQTG7F9Z 3 claims
This is likely a documentary digitization image rather than a consumer scan, made with a repro station using registration targets.?? Visible fiducial targets, orientation arrow, and the 'ScanStation' label indicate controlled capture equipment. ai:speculation · 1%
postal destination on address side?? A place name is likely present in the handwritten address, but it cannot be read from the thumbnail. dcterms:spatial · 1%
The stamp appears to be a U.S. postage stamp, suggesting the postcard was mailed within or from the United States.?? The red stamp in the upper right resembles a U.S. definitive/profile stamp format, but the thumbnail is too small to confirm country or denomination. ai:speculation · 0%
Subject: image:01KQJA2XDVE9WAR99MFQTG7F9Z 0 claims
No claims recorded yet for this subject.
closed — No more image calls. Item workflow consumes image claims/text/metadata for catalog-level analysis.
TransitionDescription
upload async Pushes realPath (the ai-tools crop) to S3 as the canonical image. When defaults.full_frame_key is present, also archives the original mat under originals/ on S3 for re-processing / QR re-extraction during Python dev. The full-frame archive is planned to drop for real-client tenants.
Available from: triaged
external_upload Sets defaults["storage"]["mode"] = "external_url". Used by IIIF imports so the image remains accessible via its external URL and subsequent transitions can run against it directly. Skips triage — the UPLOADED place lets enrich run against the remote URL.
Available from: new
triage async When fired from NEW: local-file mode. ai-tools reads defaults.full_frame_key + defaults.sidecar_key from the shared dir, produces the canonical crop + thumbnail in derivatives/, and returns deskew/type/detections. realPath is repointed at the crop. When re-run from TRIAGED: same, idempotent. Non-URL mode is the only path browser captures take. Failures keep the image in NEW (async retry).
Available from: new, triaged
enrich async Single cheap AI call (~$0.0004) against a ≤512px thumbnail, with triage OCR + flags in context. Populates title/description/subjects/people/places/dense_summary and text-type flags. Can run from TRIAGED (using ssai image_view via imgProxy, skipping S3 upload) or from UPLOADED. Re-runnable from enriched for debugging.
Available from: triaged, uploaded, enriched
extract async The optional second image call. Runs only when enrich_from_thumbnail sets pixelsDone=false. Dispatches by extraction goal/signals: filled forms → ExtractMetadata; handwriting → TranscribeHandwriting; dense print/layout → OcrMistral. Re-runnable.
Available from: enriched, extracted
close Bookkeeping only. Requires pixelsDone=true. The item workflow performs text-only catalog analysis across all images.
ignore Operator triage from the narrate screen. Hides the image from narration immediately and queues it for permanent deletion at intake close. No AI transition lists IGNORED in its `from`, so an ignored image can never be sent for enrichment -- the workflow itself is the guard, no separate check needed.
Subject: image:01KQJA2XDVE9WAR99MFQTG7F9Z 24 claims
Predicate Value Conf. Basis
ai:denseSummary
Thumbnail shows the back of a divided postcard mounted in a ScanStation capture setup, with a red postage stamp, possible postmark, handwritten message on the left, and a handwritten address on the right; the text and postal details require high-resolution review for reliable transcription.
1% —
ai:handwrittenText yes 1% —
ai:hasText yes 1% —
ai:highResGoal
read_handwriting
1%
Loss of sender/recipient names, place names, message content, postal markings, and possible dating evidence visible only in the handwriting and stamp/postmark.
ai:highResTarget
handwritten message on left half of postcard
1%
High-resolution target from thumbnail enrich
ai:highResTarget
address lines on right half
1%
High-resolution target from thumbnail enrich
ai:highResTarget
printed postcard heading and publisher or printer text along edges
1%
High-resolution target from thumbnail enrich
ai:highResTarget
postmark and stamp details
1%
High-resolution target from thumbnail enrich
ai:isFilledForm —
ai:isForm —
ai:pixelsDone
Thumbnail clearly shows a postcard back with substantial handwriting and some printed text, but the writing is too small to read reliably.
ai:speculation
The stamp appears to be a U.S. postage stamp, suggesting the postcard was mailed within or from the United States.
0%
The red stamp in the upper right resembles a U.S. definitive/profile stamp format, but the thumbnail is too small to confirm country or denomination.
ai:speculation
This is likely a documentary digitization image rather than a consumer scan, made with a repro station using registration targets.
1%
Visible fiducial targets, orientation arrow, and the 'ScanStation' label indicate controlled capture equipment.
ai:typedText yes 1% —
dcterms:spatial postal destination on address side 1%
A place name is likely present in the handwritten address, but it cannot be read from the thumbnail.
dcterms:type
postcard
1% —
Predicate Value Conf. Basis
ai:documentLikely yes 1%
ai-tools heuristic on scan.storage:marac/003/0046.jpg
ai:hasText
Tesseract via ai-tools (6 words, mean conf 34.3)
ai:peopleLikely
ai-tools heuristic on scan.storage:marac/003/0046.jpg
ai:photoLikely
ai-tools heuristic on scan.storage:marac/003/0046.jpg
ai:photoWithTextLikely
ai-tools heuristic on scan.storage:marac/003/0046.jpg
dcterms:type
tag_or_label
5%
found 3 machine-readable code(s); small text footprint matches labels or tags
dcterms:type
document
1%
visual text regions suggest document-like content
dcterms:type
text_page
1%
visual text regions suggest text blocks