OCR!
- Thread starter mikelove
- Start date
sui.generis
探花
ipsi said:
Just swung by to second that. I left my impression of pleco in the comments there as well for the benefit of the curious.
This looks spectacular. I can think of many uses for this. Is there anything like this on the market, other chinese character OCR apps ? This is something really groundbreaking that I can show even to non-chinese learners and they'll be impressed.
Thanks again and please take your time in finalizing this feature, i'll buy it the minute it's out.
Thanks again and please take your time in finalizing this feature, i'll buy it the minute it's out.
Will it be possible to take a picture and then perform the OCR off of that photo ? because that would be easier than trying to keep a steady focus on the material. For instance, if i take a picture of a sign in china, I don't want to stand in front of the sign for 5 minutes trying to decipher everything, i'd rather take a picture, then pan and zoom on individual characters.
ipsi / sui.generis - thank you! Very exciting, yes - just hope people are as happy with the finished product 
A few offline / take-a-photo ones, but nothing live as far as we know. Though as some have pointed out, pen scanner devices that you can drag across Chinese characters to recognize them have been around for a while.
The system supports that, but it'll probably only be in the app as an "experimental" feature if we include it at all - it works pretty well, but it's finicky enough (easy enough to throw off with a bent page / uneven lighting / etc) that we don't think users are going to be satisfied with the results if we make it a central part of the app.
dustpuppy said:
A few offline / take-a-photo ones, but nothing live as far as we know. Though as some have pointed out, pen scanner devices that you can drag across Chinese characters to recognize them have been around for a while.
dustpuppy said:
The system supports that, but it'll probably only be in the app as an "experimental" feature if we include it at all - it works pretty well, but it's finicky enough (easy enough to throw off with a bent page / uneven lighting / etc) that we don't think users are going to be satisfied with the results if we make it a central part of the app.
Hey mike! I really love the new OCR feature in the 2.2 release but other than chinese, could this add-on as well translate english into chinese in the meantime if E-C dicts were installed? It would be great to see "live" english-chinese translate that I can show them to my chinese friends.
anthony said:
Not at the moment - the recognizer does actually support very limited English OCR, but there'd be a lot of coding needed on our end to make it work well and it's not really a big focus for us - I suspect that within a few months the Chinese dictionary app makers like Kingsoft will have added E-C OCR lookups to their apps and at a much lower cost than our system.
character
状元
Mike, what's the speed of startup of the OCR? Is it like other camera apps (a few second delay before getting a camera feed on the screen)? I wonder if there will need to be some way to keep the camera active if one is looking up a character every few minutes (reading a bit above one's level, for ex).
character said:
Only about a second or two on the iPhone 4 at least, though it may take a couple more seconds to get it lined up / focused correctly. It fires up almost instantly when you resume from sleep - I'd say it's under a second in that case - so all-in-all you still end up with a lookup that's quite a bit bit faster than handwriting even when you're starting with a device that's turned off.
dustpuppy said:
Sometime in September but we're not being too specific beyond that - it's a brand new API and there's definitely the possibility that we might encounter a major bug we're not aware of, or that Apple might take issue with something in our implementation and require that we change / resubmit it.
dustpuppy said:
Well fortunately I don't think it's very likely - they've let all of our updates sail through so far, and in this case it's actually a pretty cool showcase app for the iPhone 4 in general
New idea we've been playing around with: a paid "Pleco Chinese Scanner" app, basically a standalone version of the OCR module with CC-CEDICT. Mostly so that people who see our cool new OCR feature can get it easily without having to fiddle around with the Add-ons screen or In-App Purchases. It would probably sell for a bit less than the OCR module for the full dictionary product, but if we can work out the licensing for this you could pay the difference in price through an In-App Purchase to activate it in the full Pleco product later on. (couldn't go the other way, though)
Upsides: easier access to OCR, usable for people without any internet connections on their iPods / iPads, promotion for our dictionary, possibly an improvement in our iTunes search ranking if they do in fact reward you for multiple apps
Downsides: confusion, negative reviews from people who didn't try it first, and it would be rampantly and viciously pirated within about 15 minutes after we released it (since we'd be relying on Apple's DRM instead of our own) - I suppose if we released it right after OS 4.1 came out and required that as a minimum version we might at least get a piracy-free window of a few weeks before OS 4.1 was jailbroken, but it would be open season after that. We probably could at least prevent the pirated OCR licenses from being moved over to the dictionary, though.
I think the Chinese Scanner idea is a great idea. Is piracy on the iPhone that bad? I can see how people in China would like that app, and I guess it's a wild west of piracy over there.
Hopefully the next iPod Touches and iPads have back-facing cameras. I think it's likely they'll have at least a front-facing camera.
Hopefully the next iPod Touches and iPads have back-facing cameras. I think it's likely they'll have at least a front-facing camera.
numble said:
It's over-reported a bit - lots of developers read things into account registration statistics that they shouldn't - but with an app like this it would be open season; expensive enough to be worth the trouble and cool enough to show off to all their friends. Though the hope is that some of them might eventually turn honest and buy the real Pleco in order to get access to more dictionaries / other features.
character
状元
Mike, do your licenses not require you to keep the interests of the licensor in mind? :wink:mikelove said:
Why not just change Pleco for a while to prompt new users to buy OCR (using a popup) and take them to the Add-ons screen with OCR already selected? Perhaps purchasing OCR could be made even easier with a special screen for it.mikelove said:
character said:
Not in such a way as to require us not to release an app for a particular platform simply because of its piracy potential, no - everyone pretty much accepts piracy as a fact of life. And I suppose one extra benefit of this is that if they can easily pirate it they might not try so hard to pirate our Chinese dictionary app, since they wouldn't need to to get access to that particular cool feature, and that ultimately benefits all of our licensors.
character said:
That seems like it might be overkill, though we could certainly move OCR up to the top slot in the Add-on catalog at least - put people just three taps away from purchasing it.
character
状元
FYI, a (non-Chinese) OCR app for iOS (and Mac): http://www.tuaw.com/2010/08/21/prizmo-i ... -and-more/
"Highly accurate OCR technology in 10 languages: English, French, German, Dutch, Italian, Spanish, Portuguese, Danish, Norwegian and Swedish."
ETA: Prizmo needs fairly bright light to work, but it does work on the plain English text I tried it on. Interesting UI with lots of ways of improving the captured image (doesn't work 'live') for better OCRing.
"Highly accurate OCR technology in 10 languages: English, French, German, Dutch, Italian, Spanish, Portuguese, Danish, Norwegian and Swedish."
ETA: Prizmo needs fairly bright light to work, but it does work on the plain English text I tried it on. Interesting UI with lots of ways of improving the captured image (doesn't work 'live') for better OCRing.