Forum Discussion
EXIF Data in JPEGs - discoverability by Delve
Has anyone had any success or know of a way that the O365 searching and indexing can include the EXIF data stored within JPEGs.
Delve is so good at returning/discoverin content from text based documents - I'd love it to be able to operate on some of the standard metadata in graphic files. At the moment my returns appear to be based on filename.jpg only ;-(
4 Replies
Hi Chris,
We consistently review what is the right amount of metadata we pull into the search service. I'll send it as feedback to the team about image files. I do beleive much of the metadata gets pulled into the system, though not certain how much applies to the Microsoft Graph relevant component. Image files do flow into Delve, but primarily due to the graph relations between people and topics, so not certain how much the EXIF data would apply in enhancing discoverability through people relationships.
Thanks,
Mark
Thanks for the response Mark. To paint a picture (exuse the pun) of what I am trying to achieve....
1. We are a School: We have around 100K documents uploaded within the tenant.
2. I have a Sharepoint site collection named 'HistoricalArchives' into which I will create subsites and Picture Libraries, eg Sport >> Athletics We are developing an ontology/classification structure at the moment.
3. I have a scanned JPG from 1959 of 4 young sportspeople at a Triathlon, we have names for 3 of those Alumni.
4. We have developed a naming convention for files, hence, name the JPEG something like SPAT-IM-Triathlon-1959.JPG
5. Using an application like IrfanView we edit the EXIF data for that JPEG and include additional detail in the following - XPTitle, XPComment, XPKeywords, XPSubject.
Throw a few thousand of these online, and imagine how great it would be to have Delve interogate those XP metadata entries.
Over the last 10 years I have used a similar system with MS PhotoGallery pointed at a network drive. Using the free Text Search Photogallery returns small accurate collections of images. It's great.
Regards.. Chris
Thx for providing more scenario details. I do get the intention. While we may not scrape this from the file itself, it is certainly possible to enhance the metadata you enter, but instead of directly into the file itself, rather to create unique columns in the library you upload the images into. You can then have a simple form (one gets created off of the columns you describe) that you add in the details; especially people's names, you can do this straight from the GAL, or as text if they aren't in your AAD instance.
I know this does not help previous uploads, but if you establish this process, where you add the details to the library upon upload instead of manually into the file itself, all the data you enter will be crawled by the search crawler and add to the relevancy data that the Graph gets built around.
Hope that helps,
Mark
