Most small-town historical societies have a back room and in the back room there is almost always a shoebox. Inside are photographs with nothing written on them and the one person who knew who was in them died years ago. Interest in AI in cultural preservation comes from a simple fact. Software can now look at a picture and say what is in it and that is more than most volunteers have time to do by hand.
Image Source:magnific.com
What image to text AI does
Image to text in artificial intelligence works by describing pictures in words. You upload a photo and within a few seconds, there is a short paragraph, something like two men beside a horse cart in front of a brick building and a street seller sitting around the corner. With image to text AI, writing inside the picture gets read as well. That covers shop signs and it covers pencil notes on the back of a print or the lines of an old letter. At present, this cannot be called perfect as AI evolves but what actually matters is that the description exists. That’s important because a photograph with words attached can be found in a search and a photograph without them can't.
Old photographs
Say a society has 600 photographs to get through. One volunteer working by hand might need months and every print would be picked up and put down again, which wears paper out. New technology can help preserve culture by doing the first pass, so each photo only has to be scanned once.
The software writes down what it sees, such as the clothing, the street and the buildings. Then somebody who grew up in town reads the descriptions and corrects them. It can't know the bakery on the corner closed in 1974. That knowledge belongs to the people who lived there.
Objects and monuments
Most of the museums around the world usually show a small part of what they own. Much of the rest has never been properly described. Think of a farm tool, a wedding dress or a church bell. With a clear photo of each, AI technology can write down the shape, the material and any decoration. That gives the catalog entry a first draft, whereas before, it might have stayed empty for years.
Monuments and old buildings are handled in exactly the same fashion. Photos taken ten years apart can be described side by side and that makes it easier to analyze where a wall has cracked or a statue has darkened compared to their previous form.
Image Source:magnific.com
Traditional artworks
Taking AI and culture together takes care, though. Models learn mostly from what is on the internet and that leans toward Western art, so local symbols often get misread. A few things are sacred and not everyone is meant to see them. So with text to image AI, patterns and carvings that once made sense to an older community that made them, get easy-to-understand descriptions and a broader coverage.
Libraries and museums
A collection is only useful if people can find things in it and written descriptions make that possible. Typing ‘wooden bridge’ or ‘local people celebrating’ brings up photographs that nobody ever labeled with those words. A good description covers:
-
Who or what is visible in the image
-
Where it was taken and roughly when
-
Any text or inscription
-
The condition of the item
-
Search words a visitor might try
Screen readers can read these descriptions aloud. This makes users who are blind or have low vision get something an untitled photograph never provided earlier.
Local communities
In a lot of places, heritage is held by a handful of people and the work of sharing it falls to whoever has spare time. A historical society can post a photo with a proper caption on its website and a student writing a school project can find it.
Language matters too. In Indonesia, people searching for gambar artificial intelligence are using the Indonesian word for image and tools that describe pictures in local languages reach many people an English-only tool would miss.
Conclusion
A description takes the software a few seconds and sometimes it is wrong, so a person should always read it before it goes into the record. That is a small price. A town that spends one winter describing its photographs will have a record that lasts well beyond the people who remember them today.
