What's new

Optical Character Recognition Review

Hello!
Based on my previous experience trying various OCR (Optical Character Recognition). I'd like to report my findings here for other users to benefit from.
This is the image file I'm using, from the Phoenix Wright game:

While it only has two kanji to identify and no katakana, the hiragana should be identified correctly.
Convertio:
upload_2018-6-24_18-57-33.png

*Convertio limits their users to 10 translations without registering.
*It's easy to use with simply selecting the image file to be analyzed.
*Once done, a Word document needs to be downloaded to find out the result.
This is the result I got:
upload_2018-6-24_19-0-19.png

So the website confused 「ろ」with 「う」. Although it's a single mistake, it would discourage me from using that resource any further as more mistakes could slip in the future.
i2OCR
*Doesn't require registration
*Doesn't limit usage.
*Is easy to use.
This is the result I got:
upload_2018-6-24_19-5-2.png

*So the sentence order is reversed. This isn't dramatic in my opinion however.
*「う」has been identified accurately.
*Both kanji have been identified correctly.
*Unfortunately, 「い」 has been identified as 「し」. It seems like hiragana is harder to identify through OCR than actual Kanji.
Free Online OCR
Doesn't work at all.

OCR Convert
*Very easy to use.
*Requires downloading a .txt document.
*No limitations
This is the result I got:
upload_2018-6-24_19-16-47.png

Apart from the missing punctuation and other glitches, the rest of the recognition is accurate although the 「い」seems on the small side.

Conclusion:
Out of the 4 websites I tried, 3 were accurate in their recognition of the kanji and only one was (almost) completely accurate. Not being able to identify the kana is a minor problem as long as it doesn't imply that kanji will also tend to be misidentified as well. I probably should have used an excerpt with more Kanji to have a better idea about this.
This being said, I'm not sure of how fast the process is compared to looking up a kanji by its header considering a connection is required as well as downloading a text file and subsequently opening it...
Thanks for reading!

Comments

There are no comments to display.

Journal entry information

Author
Zizka
Views
227
Last update

More entries in Blogroll

More entries from Zizka

Top