Talk:Wikimedia Foundation GLAM team

From Meta, a Wikimedia project coordination wiki
Jump to navigation Jump to search

Describing photographs from heritage collections on Wikimedia Commons[edit]

Hi, I would be particularly interested in working towards a harmonized approach to dealing with the description of photographs from heritage collections. See for example the discussion here. We are still quite far from having a shared vision on how to move forward, and even less so from having appropriate user interfaces in place on Commons to deal with copies and derivatives of the same photograph, e.g. photos that were uploaded in different versions (TIFF and JPG) or photos that were originally uploaded with a frame and were later stripped of the frame - sometimes resulting in a new version of the same photograph and sometimes resulting in a separate file on Commons, which should obviously share the "depicts" statements and a lot of further information with the file that was uploaded originally. --Beat Estermann (talk) 11:50, 19 September 2020 (UTC)Reply[reply]

Use of IIIF[edit]

I am also interested in learning more about how to feed IIIF compatible information into Wikimedia Commons to qualify depcits statements (e.g. indicating a box in which the depicted object can be found). In my view, this issue should be addressed in combination with the issue mentioned above (how to deal with multiple versions of the same photo on Wikimedia Commons)? - There are photographs with a frame that is often cropped away before the image is being used in Wikipedia. There are cases of scans of photo albums with several photographs on the same scan - they would typically need to be separated before applying "depicts" statements; but the link to the original scan should not be lost in the process. We need a clear vision how to model these items on Wikimedia Commons (or on Wikidata) and how to put them to use (in Wikipedia, etc.). In my understanding, IIIF would allow us to store each image only once and to create new versions of the image (cropped or separated, specific extracts, etc.) dynamically (on the fly or cached). - How do we envision the way forward in this area? How do we deal with the attached metadata (some of which would apply to all versions of the image, and some of which only to some of them)?

I am asking with a very specific use case in mind: Students of mine are going to develop a machine learning algorithm that extracts entities (object recognition) from images on Wikimedia Commons and links them to items on Wikidata using "depicts" statements. In the course of a first pilot project last spring, we ran into the above mentioned limitations on Wikimedia Commons. The students originally planned to transpose all the structured information from the infoboxes to statements, which would allow querying it through the SPARQL endpoint. This plan had to be dropped due to the lack of clarity and shared vision among the (Structured) Commons community. In my view, these limitations are rather serious, hamper current activities, and should therefore be addressed with high priority. --Beat Estermann (talk) 11:50, 19 September 2020 (UTC)Reply[reply]