OCR of Screenshots

[Shun]

Hi,

I'm still using the OCR that comes with Pleco 3.2.14, not the beta-test version, and since there can be no Screen Reader feature on iOS, I tried how taking screenshots of WeChat sessions and OCR'ing these works. Might the upcoming improvements to the OCR engine also help in situations such as the following?



Here, the OCR engine tries to recognize the background instead of the chat bubbles.

Thanks,

Shun

 

[Ashley Ben]

Quick fix for this problem, I do this a lot when I'm ordering things from Taobao and it won't let me copy and paste the text.

I resize the green-outlined box just around the text I want to OCR and it seems to eliminate most of these problems.

Not doing this in the first place would be great! But the above is a means to an end.

 

[Shun]

Of course, thanks for reminding me. But it should definitely be possible for the engine to single out Chinese characters beforehand, so let's see if it progresses to that level.

 

[mikelove]

It's probably concluding the image is white-on-black instead of black-on-white; tap on the 'photo' icon, then on the rightmost icon (the one with the two ⼻'s) and that will force it to interpret the image as black-on-white.

 

[Shun]

Thanks, that gets us to about 80% of all characters, which is respectable. Should've figured that one out as well, though.

 

[mikelove]

Also, if you email / PM me your Registration ID I can add you to the beta of our new recognition engine, which greatly improves performance in cases like this.

 

[Shun]

Thanks, I will gladly do that!