Announcing: Eye [extensible OCR tool]
category: code [glöplog]
Hot new vibrator-made project. Java 6. Very enjoyable.
This is now advanced enough to announce properly (it actually does stuff that is kind of useful). I made this within a little over two weeks. Keeps growing like madness.
Eye homepage
download (any OS)
screens
give me money (includes political commentary)
vid tutorial 1: how to build eye
vid tutorial 2: how to write your own recognizer (includes audio commentary)
- intelligent minds welcome -
(I can't promise I'll be reading the answers in this thread - I rather have a habit of not reading Pouet replies -, so if you want to contact me, you might want to use some of the other available networks.)
This is now advanced enough to announce properly (it actually does stuff that is kind of useful). I made this within a little over two weeks. Keeps growing like madness.
Eye homepage
download (any OS)
screens
give me money (includes political commentary)
vid tutorial 1: how to build eye
vid tutorial 2: how to write your own recognizer (includes audio commentary)
- intelligent minds welcome -
(I can't promise I'll be reading the answers in this thread - I rather have a habit of not reading Pouet replies -, so if you want to contact me, you might want to use some of the other available networks.)
OCR apps are so 1997
what Rowley Birkin said.
it would be nice if it could run in real-time from a live input stream
what bdk said.
O noes, it's coffee in the app icon. run!
bdk, that should be doable if you have enough processing power. however, i bet it would demand proper lighting in your room.
i have heard of iphone apps for banks that recognized checks and payment forms via the phones camera. that's awesome use of ocr technology :-)
i have heard of iphone apps for banks that recognized checks and payment forms via the phones camera. that's awesome use of ocr technology :-)
btw. i want ocr on buildings - i.e. take a picture of a house and the software will google up the location, residents, owners, bars, cafes, offices, etc.
next i want a gps that works indoors and a system to scan building structures so that the gps can locate me in a specific room on a specific floor.
go invent!
next i want a gps that works indoors and a system to scan building structures so that the gps can locate me in a specific room on a specific floor.
go invent!
Rasmus: Actually i am in a development project that is researching Inertial Navigation based on accelerometers and gyroscopes, so one can be tracked inside a building. So just wait one or two years :D
Quote:
Donations
No one has donated to Eye (eyeocr)
You could be the first donor!
could really do with a PayPal button.
Quote:
Incidentally, the whole ridiculous world of money is going away anyway - and soon.
Incidentally there are people telling this since 1900s...
and that's augmented reality, not OCR
maali: it all depends on vague fancy definitions - ocr could be a part of augmented reality or ubiquitous computing
flash: doesn't that depend heavily on delta factors? i.e. you need to know where you start without loosing any intermediate updates?
flash: doesn't that depend heavily on delta factors? i.e. you need to know where you start without loosing any intermediate updates?
@flashlight -- i've got the same idea: use inertial navigation software from the 1970's as short-term, sgort-range backup for GPS. it should be doable using solid-state accelerometers designed for firing airbags. And some trial+error refinement.
And I've got the idea, when driving through the tunnels of Stockholm on my last company trip.
And I've got the idea, when driving through the tunnels of Stockholm on my last company trip.
Quote:
Rasmus: Actually i am in a development project that is researching Inertial Navigation based on accelerometers and gyroscopes, so one can be tracked inside a building.
Casio has a camera that does that.
I'd say he looks quite far behind, is it a disability aid? :) Seriously, it's kind of cool technology but beyond that, urgh.
Well, problem currently is that with solid state accelerometers is that in a short time there is just to much error in the final position data that it will have a deviation of more then 2 meters. Atleast thats our problem at the moment :D
i once recorded an hour of acceleration data coming from the iphone on my daily train commute. no way. just noise. a perfect noise. except for few spikes here and there when the train was braking.
not to mention gyroscopes drift.
rmeht: I guess it "should" show periods of acceleration, then long periods of noise (when the train is cruising, there is no real acceleration, just some bouncing about), then some deceleration. If you could average out the data to remove the noise, perhaps it's possible to determine that there's low acceleration for x seconds, resulting in a speed of y etc.
Maybe with lots of cunning, it's possible to show at least that you started moving, but I guess the inaccuracy for something like that will be so bad it would be pretty worthless.
With gyroscope + accelerometer, it should be quite a bit better as you can tell better what the real motion is, if it's movement or rotation. Not got an iphone 4 to try it yet though :(
Maybe with lots of cunning, it's possible to show at least that you started moving, but I guess the inaccuracy for something like that will be so bad it would be pretty worthless.
With gyroscope + accelerometer, it should be quite a bit better as you can tell better what the real motion is, if it's movement or rotation. Not got an iphone 4 to try it yet though :(
psonice: the problem is that the magnitude of the train acceleration and the noise coming from the iphone sensor are comparable. plus, the noise tends to bias towards some direction that's not fixed but changing apparently in a random fashion. in the end, no serious data can be gathered, apart from a vague sense of "yes, we are moving". don't know if newer iphones are better with that, bu i doubt they changed the accelerometer just for the sake of us poor developers striving for a cheap inertial positioning system in our pocket :)