h a l f b a k e r yIf you need to ask, you can't afford it.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
1. Take a snapshot of your DVD shelves.
2. Upload to Internet Service. (this idea).
3. Internet Service: uses OCR (and related techs) to provide an
inventory of the DVDs on your shelf, which you get back as a
text file.
4. And only so it can any chance of being self-funding: make
recommendations
(like Amazon) about future purchases and
allow you to buy them.
Could also of course be used for CDs (remember them?) and
books.....
Might be useful for insurance purposes as well.
OCR
http://en.wikipedia...aracter_recognition
Wikipedia entry for OCR [monojohnny, Oct 25 2008]
Brad Templeton has the same idea for bookshelves (2007)
http://ideas.4brad....atalog-your-library
[jutta, Oct 25 2008]
Character Recognition in Bookshelf Images Using Context-Based Image Templates (Japan, 1999)
http://portal.acm.o...ation.cfm?id=840429
[jutta, Oct 25 2008]
Please log in.
If you're not logged in,
you can see what this page
looks like, but you will
not be able to add anything.
Annotation:
|
| |
Add sorting assistance with Dewey, LOC, Alphabetical and ISBN support and you might get a bun from a librarian. |
|
| |
Disclaimer: I am not a librarian. I'd bun still though. |
|
| |
For me, this would require some kind of sorting robot to fish the DVDs out from the various crevices they're in all over the house before it could be done. I think it'd be easier for there to be something like Gracenote which catalogues them when they go in the player for the first time, then have some kind of massive CD changer style device they disappear into forever, so they can't get lost. |
|
| |
I like it. Text/image recognition would only go so far, but other than that this could be quite nifty. |
|
| |
// Again, HDD space is so cheap now that it's more economical to keep media on a big hard drive than on DVDs. Why would you bother?// |
|
| |
[ Last Ditch Sales Attempt Follows.. :-) ]
Fair point of course: but I for one still have DVDs which I bought before HDD was so cheap - in fact this system could be used (perhaps...) to go and fetch the recognized movies from an internet service - allowing you to replace with HDD with minimal effort. |
|
| |
most DVDs have some kind of barcode on them, often wrapped around the hole. probably this is already a unique ID. If not, a gracenote type system plus a changer (200+ disc DVD changers are long since baked) seems best. photo-> OCR is so roundabout |
|
| |
//most DVDs have some kind of barcode on them...OCR is so roundabout// |
|
| |
True enough, but the barcode is on the back, not the side: and I think most people stack their DVDs like books, so you wouldn't see the back.... |
|
| |
I agree is OCR is fuzzy and rounadbout as you put it - but actually maybe its not so bad for recognition software: |
|
| |
1. DVD cases are (fairly) uniform - some box-sets are a bit whacky I'll admit - but a single DVD package is generally the same dimensions exactly. |
|
| |
2. There is a limited search-field here - the database (IMBD?) is finite and well-known , so this would surely help a lot with accuracy... |
|
| |
A) Have you any idea of the quality of "snapshots"? B) What software is used to read those titles to get the text file? C) Where do you upload this file so Amazon can read it?
This is a dream of someone with no idea of how computers really work. Check with your kids for useful info. |
|
| |
Rather than trying to OCR the text, it could reference the spine art, color, text and all against a database. |
|
| |
//This is a dream of someone with no idea of how
computers really work// |
|
| |
//automagical sharpening software// |
|
| |
With respect, these comments underestimate current
capabilities of OCR software... The police have automatic
number plate recognition software as part of their
standard kit now for instance... |
|
| |
Granted the point about non-standard fonts...but I go back
to my previous comment - the DVD cases are "well-known",
so you can take other cues from the case: so we are not
looking for any random variation - we are looking to
_match_ an existing title... |
|
| |
And to answer specific questions [MauiChuck] |
|
| |
A) Have you any idea of the quality of "snapshots"?
- Pretty good I would say; Don't know what unit of 'quality'
you were after here - but as for as 'resolution' goes my
Canon Ixus (5 years old) is 4 megapixel for instance: pretty
adequate I would say. (newer cameras : double-that) |
|
| |
B) What software is used to read those titles to get the
text file? |
|
| |
- urmm..that would be the software I'm suggesting here as
part of this idea. As I indicated, it would be OCR-based
(That's "Optical Character Recognition", to save you a
google-search). You realise what the 'halfbakery' is right?
Check with YOUR kids if unsure :-p |
|
| |
C) Where do you upload this file so Amazon can read it? |
|
| |
eh? You can't. I think you are just being obtuse now... :-)
This is an idea for half-bakery....I only said '*like* Amazon' ?
?? |
|
| |
To clarify, somebody would need to write what is known as
a 'computer program' which could process the 'text file' and
then interact with a back-end 'relational database' to
perform the recommendations-part of the process. This
service would probably be exposed over HTTP/S to allow a
user to interact with the whole system. |
|
| |
In other words a fairly standard 'web application': no big
deal... |
|
| |
I got carried away in my self-righteous sarcastic response
there :-) |
|
| |
Of course you upload the _image_ to the web service, the
back-end program would process the image (OCR etc) and
then produce the text-file - or then go onto interact with
the recommendations (etc) service... |
|
| |
// it could reference the spine art, color, text and all against
a database// |
|
| |
Indeed, I agree completely: the s/w could take multiple
clues here. Some aspects of the art-work are very well
defined;
the age-classification badge the 'dvd' (etc) logo, studio logo
etc. PS2, X-Box boxes (same size) are particularly uniform as
an aside as well... |
|
| |
This is a good idea and technically possible, although this kind of free-form OCR might be a little bit harder than you think. (But it's definitely easier than people who accuse you of relying on magic think, too.) I'm used to thinking about that in connection with book shelves and libraries; there's a little bit of research on that from a few years ago. |
|
| |
// Shelley Volante font.// ...edge-case though surely :)
(no pun intended). |
|
| |
Actually, I just took a picture of my DVD shelf and you
could be right about the need for some 'Blade-Runner'
Style edge-sharpening...quite a lot of glare from the
plastic casing from the flash... |
|
| |
Interestingly though a 4 megapixel picture (1600 x 1200)
taken from no more than 6 foot away , is easily able to
capture 100+ DVDs and make them (mostly) readable, to a
human at least... |
|
| |
Except 'Red Dwarf' which has some weird over-sized font-
thing going on and RIcky Gervais' 'Politics' which has
different coloured fonts... |
|
| |
Thanks for the links [Jutta], so I guess then this is in fact
baked ? (or at least not original...) :( |
|
| |
A large portion of this has been done: see http://www.readerware.com/ |
|
| |
Start with a PC+webcam or camera-equipped PocketPC. For each DVD, show the UPC-code to the video camera; the program will automatically recognize the UPC code and populate your disk-inventory database with disk name / format / actors / cover scans etc. |
|
| |
Then we just need a script to use the database contents to look for Amazon recommendations.
Probably wouldn't be hard to write a freeware version of this in Python. |
|
| |
like i said, UPC/barcode. |
|
| |
I was thinking that with the DVD
changer + software + gracenote -type DB the scanning each
dvd thing would be as easy as loading them into a changer
and pressing "fetch titles"
especially since quite a few DVD's have their volume name
set to the DVD title. (go stick one in your computer.
Does it come up in "my computer" as D:"DVD" or as D:"X-
Men II"(or whatever.)?
(on a mac, omit the D:) |
|
| |
even barcode scanning seems mostly excessive.
simply generate an MD5 sum of each disk, and search the
(hypothetical)"dvd-sum" DB on the 'net for the title which
matches your checksum |
|
| |
Heck a freeware version of this could easily be
implemented. Like this: |
|
| |
1. you sell units, and users install the software. |
|
| |
2. users load up their carousels with your dvds (depending
on how many they have, they get a different unit. 1-500
disc is already baked) |
|
| |
3. then they click "search for titles" and it goes through
the discs making a checksum for each. (this would take as
long as 15 minutes if you had a full carousel of new discs)
then, it checks with the DB for titles. For each title it
finds, it asks if it's the correct title and data. if it isn't they
submit a corrected version (their unit's serial number
would be attached, so that they can be uniquely
identified.)
then, for each disc that is not in the DB, it would ask for a
title, stars, director, producer, studio, date. If a user's
corrections and new entries are wrong more than x% of the
time, they'd lose correction privileges. they'd still have
submission permission, as only discs the system has never
seen would need a submission. |
|
| |
4. after a time, the system would become non user
editable (say, after 500000 titles or something) and would
be only editable by those unit ID's that belonged to
studios. this way, the system could be quickly set up, and
remain mostly free of vandalism. new submissions would
still be allowed for all users (after the end of the user
editable period, you would press "mark as incorrect" to
have a trusted user (studio, usually) correct bad entrys |
|
| |