Tech for Tesco / TJAM
Alec
http://api.techfortesco.com/tjamregistration/
- last I saw they had about 50 places left; given what the Mine wants
to achieve, I am defniitely going, and I suspect a bunch of you guys
would also be interested.
Just one thing: I already put myseld down as TheMineProject, and they
are enforcing 1-per-company rules, so be creative with your
attribution. :-)
-a
Richard Marr
2009/7/16 Alec <Alec.M...@gmail.com>:
Richard Marr
richar...@gmail.com
07976 910 515
Alec Muffett
On 17 Jul 2009, at 09:18, Richard Marr wrote:
>
> I've signed up too, so you should have at least a little backup.
Yay!
While we're at it: I now have a Android phone - is that book-cover-
recognition software available for it, yet?
And (thirdly) does anyone know a free OCR software stack? I seem to
remember Harvard or somewhere, publishing one. I want to rip text out
of JPEGs and PNGs
-a
Matt Barrett
Thanks for the heads up on this, Alec. I've signed up too. I saw the
Tesco API release float past
(http://news.ycombinator.com/item?id=703244), and have been meaning to
take a closer look.
Will take a look at the latest PyMine stuff before then..
Rgds
Matt
Alec Muffett
It's a moving target, but comments are appreciated. :-)
If anyone has experience of Django, I would welcome a chat; I am
currently presuming (and would like confirmation) that if I have a
model Foo:
class Foo(models.Model):
name = models.CharField(128)
...then I hope that it is legal to something like:
x = Foo()
x.name = "this is a name"
x.save()
...to stash it in the database; the reason I am worried is that the
models (from the tutorial) seem to have no explicit __init__() method,
and I am worried I might be missing something by merely instantiating
them out of thin air and trying to save() them.
It's one of many gaps in my Django understanding of the moment. :-)
- alec
Chris Adams
Count me as another signed up Tesco Jam underminer (hoho...)
I've got a portable fujitsu scansnap that can scan receipts quickly - it needs a mac. though [1]
If we can OCR that receipt, look up the items, and cross ref against the api, then dump it in our personal mine, then we'll basically have something that will let us capture our own club card data.
Better yet...
-- take the code, scanner and turn it into a kiosk type constraption
- if we can make this generate an xml document somewhere online [2]
- encode a link to it in a QR code to display on a decenty size screen
- let people with a phone take the photo of the qr code, or follow the link from an iphone or gphone, to download the xml document
Then you have a way to basically turn any receipt into mine fodder, and a by virtue of the analogue hole, created a way to break the clubcard monopoloy on access to our own spending habits.
Does either of these ideas sound like something worth having a go at hacking on at the Tjam?
If nothing else I expect it would be a quite press friendly project if we could actually build it...
Chris Adams
[1] - http://www.fujitsu.com/us/services/computing/peripherals/scanners/scansnap/s300m.html
[2] I've left security out here, as I figure Alec knows far, than I about this, and could advise on finding an elegant solution.
--
---
I'm currently only checking my email at 9am, midday and at 4pm.
If you need a response from me urgently, please call or text my mobile, or contact me via Skype (chris.d.adams).
---
Chris Adams
Stemcel Studios
The Hub
5 Torrens Street
London
EC1V 1NQ
email: ch...@stemcel.co.uk
web: www.stemcel.co.uk
twitter:chris_d_adams
skype: chris.d.adams
mob: 07974 368 229
tel: 0207 558 8971
Alec Muffett
:-)
> I've got a portable fujitsu scansnap that can scan receipts quickly
> - it needs a mac. though [1]
Likewise; where I want to go with Mine Apps includes the following:
(in no particular order)
1) Scanner-OCR receipts -> bookkeeping, sharing purchase history, etc
2) OCR JPEG photos of receipts -> more convenient than scanner?
3) OCR JPEG photos of sunday newspaper magazine recipe pages: instant
keyword reference for recipes JPEGs stored in mine; this is a really
big interest for me :-)
4) Delicious-like bookmarking engine (should be trivial, Mine
infrastructure is 90% of way there)
5) Flickr-like photo gallery / sharing with feeds
6) Tie android badcode scanner into creating "want to buy" objects in
mine, both memorising your "wants" and opening up possibility for
"personal RFP" (the latter about which I am dubious; the former I want)
7) personal "tinyurl" engine (yet more personal data better-off not
being in a silo)
8) blog engine (requires 7)
9) geolocation aggregator - consolidate all these location-sensors
into one, provide as feed to people under *my* control.
10) android mine controller (leading on to distributed mine, v2.0,
crypto, yadda ...)
> If we can OCR that receipt, look up the items, and cross ref against
> the api, then dump it in our personal mine, then we'll basically
> have something that will let us capture our own club card data.
Yep - glad to see we're on the same page :-)
> - take the code, scanner and turn it into a kiosk type constraption
> - if we can make this generate an xml document somewhere online [2]
> - encode a link to it in a QR code to display on a decenty size screen
> - let people with a phone take the photo of the qr code, or follow
> the link from an iphone or gphone, to download the xml document
That's a *really* cute idea. Really cool. I like it. When can we
see the code? :-)
I am already playing with the idea of home-printing business cards
with individual QRcodes that relate to individual, preconfigured,
single-relationship Mine feeds or objects; that would allow tracking
of who uses what - I can give 1000 people 1000 speculative feeds, and
if/when they eventually take-up the feed seriously I can bump the
revision number and provide a new URL to improve security.
Ketan at the UI-design meeting (http://twitpic.com/9hsuf) made the
comparison to some kind of marketing software, and I can see a
parallel - except this is *ours* :-)
Also, this appeals to the geocacher in me - putting stuff out in the
wild and seeing what happens to it, throwing messages-in-bottles into
the sea.
> Then you have a way to basically turn any receipt into mine fodder,
> and a by virtue of the analogue hole, created a way to break the
> clubcard monopoloy on access to our own spending habits.
We're both on page 2, then. :-)
> Does either of these ideas sound like something worth having a go at
> hacking on at the Tjam?
Dunno about "at the T-jam" but yes; my interest for T-jam is to see
whether the API can/does provide direct access into my clubcard
purchase list, and if not, why not? Plus to get an overview of what
information is being offered and see if it can / needs to be
repurposed. The guy running the shindig so-far has been making all
the right noises, I am quite impressed.
> If nothing else I expect it would be a quite press friendly project
> if we could actually build it...
I love all the ideas so far; anyone else?
-a
Adriana
Have a problem with JPEGS of recipes or any graphical representation
of any text really. No good for sharing or deeper search. I'd rather
have the scan converted to text properly. Shouldn't be a problem, e.g.
evernotes core technology is writing and speech recognition, I
believe, so these text conversion technologies are around. (though
evernote has the patents, so we'll see what else is out there.)
Re scanning receipts - not holding my breath in terms of wider user
adoption. I have a receipt scanning application on my g1, with ability
to tag, take a photo, name it, date it etc, I believe even export it
somehow. I started to use it in earnest a few months back, imbued with
zeal for Mine! use. Alas, it's cumbersome (not the application, but
the fact of having to do it), time-consuming and I simply stopped
doing it, after much effort. The problem is once you miss a few
receipts, your data set is not complete, never mind the OCD element of
collecting your receipts. So, sorry to piss on the parade here, the
point is that even though I really really want my purchase data, I
couldn't keep up the scanning and tagging - so what are the chances of
'normal' people scanning stacks of receipts after each shopping? Also,
what about all my online purchases? Retype them to a spreadsheet? Take
a screenshot? Don't think so...
Best option is to get retailers/vendors/anyone with whom I transact,
give the data back to me - i.e. small exportable flat file(s)
somewhere along the transaction. (this doesn't solve the issue of
offline transactions, of course, so yay! for receipt scanning. :P)
That is why, as far as Tesco API is concerned, my main interest is my
data as customer - from purchase history to potential logs of browsing
on their website if applicable. However, I know that's not the point
of Tesco API, which is aimed at developers. But, I'd still like to get
Tesco give me (and other customers) back my data. More on that
later...
A xx
2009/7/17 Alec Muffett <al...@crypticide.com>:
Alec Muffett
>
> Have a problem with JPEGS of recipes or any graphical representation
> of any text really. No good for sharing or deeper search.
Agreed, it's a pain; my intention is that (knowing OCR to be flaky) to
store the JPEG as an "Item" in the Mine, and to copy-out the text to
the Item.description field where it will become searchable, and the
cook can refer back to the JPEG in cases where the scanned text is
unclear; in the full course of time I want the OCR to do a perfect
job, of course, exactly as you say.
However the above functionality will be generically useful, and can be
lashed up fairly quickly.
> Re scanning receipts - not holding my breath in terms of wider user
> adoption. I have a receipt scanning application on my g1[...]
> The problem is once you miss a few
> receipts, your data set is not complete, never mind the OCD element of
> collecting your receipts. So, sorry to piss on the parade here, the
> point is that even though I really really want my purchase data, I
> couldn't keep up the scanning and tagging - so what are the chances of
> 'normal' people scanning stacks of receipts after each shopping? Also,
> what about all my online purchases? Retype them to a spreadsheet? Take
> a screenshot? Don't think so...
Agreed; I want, but I can't rely upon that. I am too lazy.
> That is why, as far as Tesco API is concerned, my main interest is my
> data as customer - from purchase history to potential logs of browsing
> on their website if applicable. However, I know that's not the point
> of Tesco API, which is aimed at developers. But, I'd still like to get
> Tesco give me (and other customers) back my data. More on that
> later...
:-)
-a
Alec Muffett
18 places left now.
Daniel Farinha
Dan
Alec Muffett wrote:
> 18 places left now.
>
>> On Jul 16, 2009 10:53 PM, "Alec" <Alec.M...@gmail.com
Alec Muffett
On 18 Jul 2009, at 04:29, Daniel Farinha wrote:
> I don't suppose they cover transportation costs? :(
suspect not. i could fedex some ice-cream? :-)
-a
Alec Muffett
posted, so the sessions at least will be available.
if i can work out the qik android user interface before then, i will
try to do some in real time. :-)
-a
Richard Marr
> recognition software available for it, yet?
It sure is... review below:
http://www.androidapps.com/t/snaptell
I can't recommend OCR software, I've stayed clear of that so far. Anyone?
Rich
Chris Adams
He had a very early version that took receipts, OCR'd the data and spat out text.
It's possible to turn it into a restful service, but it needs calibration and training to get good at parsing out details from the receipts.
There are closed source ocr services, which may be better as well as Adobe's own reader one, but again, I do'nt know how easy they would be to interface with a web app.
Long story short - we have opensouce one that's pretty crappy, but can parse out some data, already - that might be enough for a proof of concept on the night.
C