(Archived) Canon P-150 OCR vs Evernote OCR

[jjziemba 0]

I am a premium user, and previously I using a multi-use scanner. It did a lousy job of OCR, and when uploading Evernote saw it was already done (poorly) so it didn't go any further with OCR (by design).

So, I was pretty excited to get my P-150 in the mail today. I was expecting it to do a better job with OCR, but not so much.

If I scan at 600dpi, it gets a bit better, but still missing quite a bit.

I am testing both in a PDF Viewer, as well as searching within Evernote, once it gets uploaded.

So I shut OCR off, and let Evernote take care of it. It didn't get every word, but it did a MUCH better job than the Canon software.

Am I doing something wrong, expecting too much from the software, or is the Evernote service that much better?

Are most premium users shutting OCR off on their scanners?

much thanks

=jz

[jbenson2 2,147]

I'm a premium user and have a Fujitsu ScanSnap S300. It is a miniature scanner that sits beside my monitor and OCR's both sides of all documents.

I haven't found anything missing so far - 8,000 notes with 2,000 scanned PDF's. Evernote picks up an amazing amount of very fine detail, even the small print on the backside of my bills. I actually test it occasionally by searching for some of the fine print. Not a single failure so far.

Remember that Evernote search only works on full words or partial words starting with the first letter of the word.

It will not find characters in the middle of a word.

Yes: Evernote or Everno

No: vernote or vern

[jjziemba 0]

You said "Evernote picks up..." - so are you using Evernote's OCR on the Snapscans?

The canon & fujitsu seemed very similar, so I expected similar results.

I am searching for full words. I scanned my Insurance card, then searched for "SAAB" and "Amica"

Evernote's OCR picked them up.... Canon's OCR did not.

Thats just one example.. happened on quite a few words.

[jbenson2 2,147]

Yes, I let the ScanSnap do all the OCR work before sending it to Evernote. It takes an additional 30 seconds but I know it is done.

Even though I am premium, I have tested the Evernote process and found that in some cases it takes someone on Monday morning to give the server a kick to finish up some of the documents sent over the weekend.

My car insurance policy summary is 7 pages long. I found the following:

• "PROOF OF INSURANCE CARD" is on the last page

The smallest font I could find was:

• "provisions remain unchanged"

In faint light blue color was

• "not intended to serve"

I even found these letters

• COMP COLL ERS UM UIM PIP

[jjziemba 0]

Thanks for the additional info!

I need to play with this some more. I see no difference in speed with OCR turned on, although I know it is doing it because I can search at least *some* words.

To make matters worse, if I turn OCR off on my scanner, and let Evernote do it, the searching only works from the Windows Client. On the website, none of words I search for are found. Very frustrating.

[jjziemba 0]

Dunno - scanned some more stuff, with OCR turned on, and now it is catching just about everything.

[jbenson2 2,147]

Thanks for the additional info!

I need to play with this some more. I see no difference in speed with OCR turned on, although I know it is doing it because I can search at least *some* words.

To make matters worse, if I turn OCR off on my scanner, and let Evernote do it, the searching only works from the Windows Client. On the website, none of words I search for are found. Very frustrating.

Glad to help. There should be a noticeable delay for the OCR process after the machine finishes the scan. I'm going to run some documents today without the OCR and double check the time it takes Evernote to do the OCR. Weekdays seem to be faster than weekends.

Dunno - scanned some more stuff, with OCR turned on, and now it is catching just about everything.

Great. I suggest that you continue to test the OCR results on a regular basis.

[jfkaess 28]

I have the Canon P-150 and several months ago, I turned off the OCR after using it for about a year. I had noticed that Evernote search was not returning hits I expected it to find, and after following instructions to load the pdf in an external viewer to check the OCR'd text discovered the P-150 OCR was doing a pretty bad job of getting the text right. Evernote so far is doing a better job. Neither is perfect.

[jbenson2 2,147]

The OCR process is more of an art than a science. The font style, spacing, color, resolution all affect the results.

For instance: You might get a hit for the word Evernote with Ever note, but not Ever note (1 space vs 2 spaces)