Pleco OCR: ATTN Mike Love

[studychinese]

Mike, please implement the the OCR that is used in Hanping Pro on Pleco. For me the Pleco OCR is basically unusable. 

 

Except for the Cantonese dictionary, I own just about every pleco addon that you can buy. I often end up using Hanping just because I use OCR in conjunction with my graded readers. I'd like to use Pleco. 

[vellocet]

Unusable?  In what way?  I find the Pleco OCR to be quite usable, unlike, for example, Waygo.  

[mikelove]

Sorry it’s not working well for you. If you email me (mikelove@pleco.com) with your device configuration and a description of the problems you’re having I can probably suggest some settings changes to make it work better for you.

 

We are working on two big OCR upgrades - a new camera capture system (already out in beta on iOS - “New OCR” in Settings / OCR) and a second / alternate algorithm we licensed from another provider that’s a bit slower but tends to do a better job with text detection in complex environments - but I don’t have a release date for them yet (we’ll probably finish implementing them during what I expect to be a very long public beta for 4.0, though the new algorithm at least could actually make its debut much much sooner).

[DavyJonesLocker]

@mikelove

 

It's there anyway you can use Ocr on wechat with the later versions of wechat? I remember seeing something on your forum about this.

 

[mikelove]

Screen OCR actually works fine, it’s the free Screen Reader function that doesn’t work because Tencent have started blocking accessibility services from accessing message texts. No way to fix that on our end (but maybe eventually Tencent will fix it on theirs) - only workaround is to install an older version of WeChat - but that new algorithm I mentioned should hopefully make Screen OCR significantly more reliable.

[艾墨本]

I second study Chinese. It’s not unusable but Hanpings OCR is the far and distant winner.

[mikelove]

Most of the differences we’re aware of can be addressed with tweaks to settings, but if there’s a mass of dissatisfaction out there we can try to hurry some new things up :-)

[vellocet]

Does everyone here know that Pleco's OCR only works with the standard Chinese fonts?  Are people trying to shine it on flashy advertisements and noticing it doesn't work?

[imron]

It doesn't really matter, because the point is that two posters have said Hanping Pro works in the situation they're trying to use it for and they'd like Pleco to be able to do that as well (and the OP specifically referred to using OCR on graded readers, so likely standard, printed, Chinese fonts).

 

I've never used Hanping Pro and I don't use Pleco OCR (but love everything else about it), and maybe it's just a matter of tweaking the settings, but your comment really doesn't add anything to the discussion.

 

BTW, Mike, if studychinese provides you with device information and settings and you manage to tweak the settings to fix things up, be sure to come and update this thread to tell people what tweaks they should be applying.

[mikelove]

Will do - I edited my previous post to remove a request for more details since I didn't want to turn this into Yet Another Pleco Tech Support Thread, but anybody having issues with OCR is welcome to email me or post in our forums and we'll do our best to help you get it working better.

[studychinese]

Thanks for the replies. 

 

Unless I am doing something wrong with Pleco, all I need to do with hanping is point my camera at the page, the page is then captured, and all I have to do is move the line to bring up the definitions. 

 

With Pleco I have a box to adjust, and the screen is constantly moving in my shaky hands. That for the live OCR. The still OCR doesn't seem to manage what Hanping manages. 

 

It's hard to describe the differences but there is an clear difference in usability. Everything else I like about Pleco. 

[mikelove]

Thanks.

 

Box to adjust: tap on the 'camera' icon at the top of the OCR screen, then tap on the second button from the left in the second row of buttons (box with dots on the corners) and that will put Pleco into 'crosshairs mode' where you'll no longer have to deal with a resizable box. (you might also want to turn off 'allow multiple lines' in Settings / OCR / Live Video OCR to improve accuracy a bit)

 

Capture frame: our next major update should support capturing + holding the current frame of video in live OCR (already supported in our experimental "New OCR" on iOS - also lets you flip back through the last dozen or so frames of recognized video if it froze on the wrong one). Two other options that might help in the meantime:

• Settings / OCR / Live / Configure motion detection / Button-only resume: with this enabled, once the OCR 'locks on' to text (yellow frame) it won't start recognizing again until you tap on the 'resume' button, even if your hands move.
• Settings / OCR / Live / Pulse mode: this takes a slightly different approach, it recognizes continuously while you hold down the button and stops as soon as you let go; a lot of people find that helps with shaky hands.

You could also increase the 'resume threshold' percentages in Settings / OCR / Live / Configure motion detection to increase the amount of motion required before Pleco will un-pause and start recognizing again.

 

We're hoping to have a preview version of that OCR revamp on Android pretty soon, actually (as on iOS, it'll be an off-by-default option in our release app) - there's some uncertainty over whether Google might add some additional hoops for users to access our non-OCR Screen Reader feature (they can't block it entirely but they might add extra hassle for users to enable it), and so we're prioritizing a revamp to Screen OCR which will also come with a lot of the new stuff we're doing for other OCR modes.

[艾墨本]

@mikelove

 

Have you tried the Hanping and its OCR? It's not really a matter of more features. It's more a matter of simplicity. Where Pleco, as a product, leans towards making every configuration possible, Hanping seems to learn more towards giving you exactly what you need. The OCR is a good example. I open it, and it works. I didn't have to customize it first. 

 

I say this while having purchased most all of Pleco's dictionaries and customized the app heavily. I never felt a need to customize Hanping, though, with the exception of switching the coloring of tones to match that of Pleco.

[mikelove]

I'd rate that less as a matter of simplicity and more as one of offering better defaults; if he added 30 or 40 more options to his app I don't think it would feel meaningfully more complex, having a bunch of buried settings most people don't use doesn't really hurt your product unless people can't use it well without mastering those settings. (that second bit is the biggest issue with our flashcard system, and a major reason that revamp is taking so long)

 

But we definitely need better defaults in our OCR - the current default configuration is a product of age (we launched OCR with the rest of our Android app in January '11, when the newest version of Android was 2.3 and the current Samsung Galaxy S was just called the Samsung Galaxy S because they'd only made one of them so far, so we were building around some very limited hardware specs) and user expectations; throwing a totally new configuration at people who've been using our OCR happily for years is not necessarily a recipe for satisfied customers. So they're definitely in line for a revamp, it just has to be done carefully.

 

My current thinking is that there'll be a split between a mode designed for live scanning - similar to our current UI but with our new state-of-the-art capture / motion detection code making it smooth and jitter-free - and a mode designed around 'snapshots' where we keep the capture in app (unlike our current still OCR) and low-resolution and freeze instantly on a snapshot as soon as you stop moving and we have a clear view (akin to a document scanner app), the former being more about environmental scanning (signs etc) and the latter more about documents. Haven't yet decided whether the document scanner would default to tap-on-every-character or sliding around with a crosshairs / box - if we assume people are capturing these snapshots at text sizes that they're comfortable with reading un-zoomed + we pair this with our newly improved text region detection system, the former would probably be better. But since this is a redesign we can be flexible; there'll be a nice long beta/preview period for people to send feedback.

[vellocet]

I would really like a preset to change the tone colors to red/green/yellow/blue instead of having to pick colors from a color wheel.  I don't know who invented that color scheme that uses purple, but I never use it.  

[mikelove]

That color scheme actually dates all the way back to 2002 when my college roommate came up with the idea; was based on the colors of markers he owned. But aside from inertia / familiarity I'd defend it now on the basis that purple is higher-contrast than yellow and more memorable / distinct from its neighbors than orange; since the goal here is to remember colors, if the purple is noticeable enough to bother you then it's actually kind of doing its job :-)

 

But the color wheel is indeed lazy on our part - the plan is to replace that with a palette of a dozen or so colors (coordinated to look good together, and with corresponding night mode colors attached) so that you can reconfigure your entire tone color scheme in just 5 taps. (you'll still be able to override this with exact colors, of course) Picking the colors is actually the tricky part, given the LCD/OLED split in Apple land and the sloppy maelstrom of system color profiles on Android :-)