Photograph your shelves to catalog your library


A lot of people want to catalog their extensive libraries, to be able to know what they have, to find books and even to join social sites which match you with people with similar book tastes, or even trade books with folks.

There are sites and programs to help you catalog your library, such as LibraryThing. You can do fast searches by typing in subsets of book titles. The most reliable quick way is to get a bar code scanner, like the free CueCats we were all given a decade ago, and scan the ISBN or UPC code. Several of these sites also support you taking a digital photograph of the UPC or ISBN barcode, which they will decode for you, but it's not as quick or reliable as an actual barcode scanner.

So I propose something far faster -- take a picture with a modern hi-res digital camera of your whole shelf. Light it well first, to avoid flash glare, perhaps by carrying a lamp in your hand. Colour is not that important. Take the shelves in a predictable order so picture number is a shelf number.

What you need next is some OCR of above average sophistication, since it has to deal with text in all sorts of changing fonts and sizes, some fine print and switching orientations. But it also has a simpler problem than most OCR packages because it has a database of known book titles, authors, publisher names and other tag phrases. And it even would have, after some time, a database of actual images of fully identified book spines taken by other users. There may be millions of books to consider but that's actually a much smaller space than most OCR has to deal with when it must consider arbitrary human sentences.

Even so, it won't do the OCR perfectly on many books. But that doesn't matter so much for some applications such as search for a book. Because if you want to know "Where's my copy of *The Internet Jokebook*" it only has to find the book whose text looks the most like that from a small set. It doesn't have to get all the letters right by any stretch. If it finds more than one match it can quickly show you them as images and you can figure it out right away.

If you want a detailed catalog, you can also just get the system to list only the books it could not figure out, and you can use the other techniques to reliably identify it. The easiest being looking at the image on screen and typing the name, but it could also print out those images per shelf, and send you over to get the barcode. The right software could catalog your whole library in minutes.

This would also have useful commercial application in bookstores, especially used ones, in all sorts of libraries and on corporate bookshelves.

Of course, the photograph technique is actually worthwhile without the OCR. You can still peruse such photographs pretty easily, much more easily than going down to look at books in storage boxes. And, should your library be destroyed in a fire, it's a great thing to have for insurance and replacement purposes. And it's also easy to update. If you don't always re-shelve books in the same place (who does) it is quick to re-photograph every so often, and software to figure out that one book moved from A to B is a much simpler challenge since it already has an image of the spine from before.


I don't know how the large libraries look for mis-shelved or misplaced books. Whenever I wanted such a book I wished they had a camera/device with the capabilities you described. It may also be useful for some large bookshops that don't have a good catalogue to create something very basic.

An application as you describe already exists: Delicious Library (only for Mac OS X)

That's like the other applications I described, you have to get the book out and photograph the bar codes, one at a time. A cuecat is much faster than that, it turns out, but what I describe would be fifty times faster than either.

Dang, Brad...this is a great freaking idea. Thanks a LOT for the suggestion. I've monkeyed around with the ISBN lookup/cuecat before, but I never thought this far. Funny, though...b/c I DO use photos for my readinglist...I'll just shoot a phone pic when I finish reading a book, and throw it up to Flickr so I don't forget about it. Can't believe I never thought to stretch this to inventory.

I know what I'm doing when I get home tonight! And the great thing is that it'll take about 10 minutes total (I do have some shelves with 2 layers of paperbacks, so I have to do a bit of moving). Thanks again!

Cool idea, Brad. I was thinking that you could just use something like one of those ocr pens that reads as you swipe text across it and you could swipe the titles, but your way is way better, and doable to some degree. If people can crack Captchas right and left, they should be able to crack a bookshelf.


Joshua Gulick

Yes. Such a technology is needed to search for a missing book in my friend’s independent bookstore.

Add new comment