Idea: Book recommender
Website idea.
You take a picture of your bookshelf and upload it to the site. The site OCRs all your book spines. Once it has data from enough users it can recommend books that you might like.
This is better than Amazon recommendations because it will catch all your books, not just the ones you bought on Amazon.
You can make money through affiliate links to Amazon, though.
(bookscanr.com is free as of this writing)
Matt McClure on 3 November 2009 at 12:25 am
Is your bookshelf organized by spine color? Or is that a stock image? Or lighting?
Nat Friedman on 3 November 2009 at 12:27 am
Actually one of my bookshelves is but it doesn’t look nearly that good
http://www.flickr.com/photos/natfriedman/3334231470/
The photo is from flickr.
Jered Floyd on 3 November 2009 at 1:49 am
I like the idea, although even assuming a good OCR engine that can deal with wild spines, this’ll require dozens to hundreds of photos to get enough resolution. As long as they can be easily uploaded, I guess that’s not a big deal.
Chris on 3 November 2009 at 2:19 am
I don’t think the OCR is going to work out; you just aren’t going to have enough pixels dedicated to each book, even with a good camera which most people don’t have, and current state of the art OCR needs crisp black text on a white background and so on.
The approach that seems more promising to me is using cell phone cameras to scan the barcodes of each book to build a library that way — see e.g. http://www.brendonwilson.com/blog/tag/barcode-scanner/ for a barcode-to-Shelfari client, or you can use a mix of Barcode Scanner and Clipbot on an Android phone.
– cjb.
Kelly Clowers on 3 November 2009 at 2:36 am
just use LibraryThing and a cuecat
Nat Friedman on 3 November 2009 at 2:41 am
Jered, Chris
I think a standard non-phone camera should be able to get enough information from each spine to make it identifiable. You don’t have to actually be able to read each character, but I bet you can learn to identify books by spine. There could be some kind of guideline like “30 books per picture, 5MP minimum.”
And anyway, there’s no way I am going to scan the barcodes of my 100s of books. I don’t know about you guys but I’m not sitting on a lot of afternoons where I think to myself “You know what I want to do right now? Scan barcodes for 5 hours.”
Jonathan Hernandez on 3 November 2009 at 8:31 am
Sounds good, I dont know if “Amazon Remembers” (Amazon iPhone app) has an API, it could be used for that idea, Amazon already has the recommendation engine, the problem here is the app to identify the distinct books (or maybe it can be used for all the things in your room)
Ben Maurer on 3 November 2009 at 8:36 am
Doing the OCR would be hardish. It’s probably solvable to some extent, but definitely not by off the shelf software. I think the amazon referral fees could be worth enough that it’d make sense to mturk the problem (doing text detection is much easier than accurately digitizing it). This is especially true if you progressively catalog the user’s library as they start buying books based on your referrals (so that you don’t spend tons of money on unprofitable users).
oliver on 3 November 2009 at 3:16 pm
The book spine OCRing would be really useful, just to get a digital list of one’s own book shelf. What is the current state of free OCR engines?
oliver on 3 November 2009 at 3:22 pm
PS: would it be possible to make a “movie” of the shelf, ie. moving the phone on front of the shelf and letting some software stitch the frames into a huge multi-megapixel image? Could current panorama stitching software be used for this?
Wolfgang on 3 November 2009 at 5:21 pm
Wow, that is a cool idea! There are lots of good book-communities out there which allow you to add all your books to some kind of virtual shelf (e.g. lovelybooks, goodreads, librarything). But who wants to add 100s of books manually… The coolest thing would be to integrate such an OCR-approach into one of those sides. Or maybe, if some of you are interested, we could develop such an OCR-book-scanner and see what we can do with it.
Wolfgang on 3 November 2009 at 5:57 pm
http://ideas.4brad.com/photograph-your-shelves-catalog-your-library
Io McKinnon on 3 November 2009 at 9:15 pm
Didn’t take long for someone to catch up:
Registrant:
c/o BOOKSCANR.COM
P.O. Box 821650
Vancouver, WA 98682
US
Registrar: DOTSTER
Ian McKellar on 3 November 2009 at 11:19 pm
Why stop at books, surely it can’t take that much computing (or mechanical turk) to identify and recommend stuff.
Also, it should be able to recommend living rooms to thieves.
Luis on 4 November 2009 at 2:58 pm
The recommendations side is almost certainly librarything, though I believe Amazon will now also give you recommendations on things you didn’t buy there. As for the rest, my current plan is to scan everything as I unpack it when I get out of storage at the end of the honeymoon