How to combine an object’s image and its catalog card to catalog is everyday work in a museum, a picture gallery or an art phototheque. Anyone with the piece in front of them looks at the form, the condition, what it depicts, a signature or an inscription; and consults what is written about it: the card, a book’s index of illustrations, an inventory sheet. Neither source is enough on its own. The image does not say in which year the work was made or how large it is; the card does not say how it looks.
Janium Collect works with those two inputs when it helps catalog, and the record changes when there is only one.
What each source contributes
Some data is read by looking; some data has to be declared.
From the image comes what is visible: the object type, what the work depicts, compositional detail, apparent condition, a signature or a mark when they are legible. That is descriptive information, the answer to what this is and how it looks. An inscription on the piece is transcribed from what is read, not from what one would expect to find.
From the card —or the index, or the inventory— comes the documentary data: the exact creation date, the measurements, the accession number, the declared technique, the collection that holds it, the photograph credit. Those data are not in the image. A painting does not carry the year it was painted; that an oil measures 73 by 92 centimetres is not known from looking at a reproduction, and the exact technique —whether it is oil or tempera, whether the support is canvas or panel— often cannot be told with certainty from a photo either.
In cataloging art with CDWA, the author of the work and the photographer who reproduced it are different data and do not come from the same source: who painted the picture is documentary data from the card; who took the photograph, likewise. The image is of the reproduction; the authorship data, of what is written. Collect keeps them separate in the record instead of confusing the photographer with the painter.
How they are combined in a record
They are two distinct inputs. The card or index is respected as it stands: author, title, date, technique, collection. The image of the work describes what is seen. The record joins the two: from the card comes the data that has to be declared; from the image, what is visible. Neither replaces the other, and the model does not have to «understand» the card in one pass.
There is a reason to copy the card instead of asking the language model to interpret it in one go. An index sometimes carries errors from the original —a misprinted date, a doubtful signature— and if the model corrects them while transcribing, the error disappears without a trace and with it the chance to audit it. The data is transcribed as it is; correction, if needed, is decided later on that faithful transcription.
Technique and exact date come from the card because they are not deduced from a reproduction. Completing them from the photo would present as verified a fact the image does not support. An eighteenth-century oil and a later copy can look very similar; the date that separates them is in the documentation, not in the image.
Nor is the piece identified with a known work by stylistic resemblance. A painter produces dozens of similar compositions, and the remembered title is often that of another in the series. What the piece supports —object type, what is depicted, a visible signature— goes into the record; what lives only in the model’s memory does not.
What happens when a source is missing
A record made from the image alone is partial, and Collect treats it as partial. If there is no card, the fields that depend on it —exact date, measurements, accession number, declared technique— stay empty. An empty field is pending work someone completes later from the documentation; a plausible datum with no backing enters the catalog looking correct and can sit there for years.
The consequence is that the quality of the card conditions the record. A complete, well-transcribed card produces a complete record on the documentary axis; an incomplete card, or its absence, leaves gaps on that axis. Collect still describes what the image shows and completes what the piece itself supports —a signature, an inscription, what is depicted. It does not fill with a guess what neither source declares.
The cataloger receives a record where the visible is already described, the documentary is already incorporated when it came on the card, and what is absent is marked as absent.
To continue the conversation
If in your institution digitized material lives alongside cards, indexes or inventories somewhere else, and cataloging well means joining the two, write to us at info@janium.com. We are interested in how the images and cards stand, and how complete the documentation of the holdings is.