Jump to content
  • 101

Extracting text from image (OCR)?


Michael K.

Idea

Hi,

 

a search in the forums only turned something up towards PDFs and that wasn't very helpful. I am wondering if Evernote has the ability to extract the text (via OCR) from an uploaded image? I am thinking here of something similar that Microsoft's OneNote is doing (you can right click on an image, then tell it to "copy text" and then paste that text anywhere you like)? It would be greatly helpful for indexing / working with images / screenshots of articles and actually complete my evernote wish list (well, if I'd be nitpicking I'd love to see the voice to text as well, Microsoft is sorta kinda working on that, the ability to link certain notes to certain time stamps in the audio has proven quite useful in the past, but now I am rambling / digressing).

  • Like 3
Link to comment

Recommended Posts

  • 0
  • Level 5*
34 minutes ago, Tylast2 said:

Yup...sorry I should've been more clear.  Copying/using the OCR'd data is what I'm trying to do.

No sweat.  Google Docs / Office Camera / Adobe Scan will all do that for you (I'm told) but as discussed above,  it isn't (currently) an Evernote option.  There's a major app refresh coming (again - I'm told!) which might answer your prayers,  but we'll have to wait a little longer to find out...

Link to comment
  • 0
  • Level 5

EN OCR does not work like that. It builds a search index, and that is it. This is different from an OCR that really is intended to transfer the picture of a text into a Computer text. The best known software to do this may be the Abby FineReader. AN individual license sells for 199$ at the moment. 

There was no communication by EN about changes to come.

In my setup my scanning software does its own OCR. The same it true for my handwriting app, GoodNotes 5. It creates an OCR of the handwritten text as well. The scanned or exported documents I create thus come with their own OCR embedded, and import with it into EN. EN then uses the existing OCR data, and build the search index from it. Problem solved ...

Link to comment
  • 0
  • Level 5*
3 hours ago, kblodorn said:

please make the OCR function capable of adding the note's title.

Hi.  Kind've by definition the Evernote note title is plain text.  If you're looking for some way to convert Rocketbook entries into Evernote notes,  I'd guess it ain't ever going to happen - but you need to create a separate feature request to suggest it.

If that's not what you're suggesting,  please give more information.  This thread is about copyable OCR'd content.  I think it's still the case that Evernote's server based image OCR is not copyable as text,  while searchable PDF files do contain text.

Link to comment
  • 0
  • Level 5

I could recommend a nice group of apps, capable of OCR as well. Or any newer iOS / MacOS device that gets a build in OCR for pictures by default. In fact OCRing is not that exceptional any more.

What EN does is something different, and I won't discuss if it is still up to date: They OCR exclusively to build a search index, and allow to highlight the positions where the searched string appears. And they run it on the server, not locally. To OCR handwriting it must be in a picture, not in a pdf.

If you like your Rocketbook, fine with me. If the app does OCR and send the text into EN, fine as well. 

Link to comment
  • 0
  • Level 5

It never has been a feature.

Looking at how EN does OCR, and for which uses, I doubt they could employ their current approach to do it. Their OCR is server based, and the main purpose is to build the search index. Since they integrate text and handwriting recognition, the combination is pretty unique.

Link to comment
  • 0
  • Level 5*
On 9/25/2022 at 6:34 AM, StePap said:

It feels just nonsense that I can search text in images but cannot extract it from them. I'm pretty sure it would require a short time implementing this additional feature...

Hi.  If you could just dash off the code and send it to Evernote I'm sure they'd be grateful,  but even if presented on a plate Evernote couldn't necessarily implement downloadable text immediately.  They have 1,000 other feature suggestions from users who would all complain if their favourite feature were not implemented next;  and they're probably already working on the next dozen or so features that will be implemented next and have no time to view this one.

Meantime there are any number of online and free services already providing this feature which has probably moved it down Evernote's priorities.  Plus this feature request has 90 votes as against hundreds for some other features they haven't gotten around to yet!

Link to comment
  • 0
  • Level 5
On 9/25/2022 at 7:34 AM, StePap said:

It feels just nonsense that I can search text in images but cannot extract it from them. I'm pretty sure it would require a short time implementing this additional feature...

OK, another expert joined the forum - welcome !

EN OCR is a server side function - it runs in a batch routine on the server, some time after a picture has been uploaded inside of its note. The result is a search index, keyword - note (GUID) - note element (file #) - position. So first it is not on device, and second it is not instantaneous.

If there are 10 words recognized, there are 10 entries (at least), plus several more. The additional ones are there because OCR is made for search, not for text extraction. Similar looking entries will be saved as well, so if there is „cone“ in the picture, words like „come“ and „code“, but as well „cane“ etc. will be probably saved in addition. This fuzziness serves a better search hit rate - but is not beneficial if you want to extract text.

Are you still convinced it would only take a short time implementing this additional feature ? Tell you: It would mean an entirely new text recognition module. The OCR to build the search index is useless for text extraction.

Solution: Use the text extraction tool build into your OS (iOS, Android, MacOS, not sure about windows) or get an OCR app for your device. Save the picture, extract the text.

Link to comment
  • 0
  • Level 5*
1 hour ago, Paul Olson said:

The "Copy Text from Image" capability found in Google Photos should be the model for this functionality. It is easy to copy all or some of the text that is recognized on the page. I don't want to do this in a separate application as a workaround.

Hi.  In 15 years or so I haven't ever wanted to do it at all.  Can you explain the need a little more fully?  Why would Evernote need to incur the possibly considerable cost of revising their (I assume) highly complex editor / OCR process if a free alternative already exists?

  • Like 1
Link to comment
  • 0

In the 15 years that I have been using Evernote I have wanted this functionality frequently and I find this functionality creeps into my workflow on almost a daily basis. Whether it is a document I have scanned or a screenshot I snapped, I would quite often like to extract the text from the image. Google Photos and iOS Photos app added this functionality in recent years. I don't see why this would not be relevant to others who use Evernote as a document storage solution, particularly for scanned images that flow directly into my Evernote application.

  • Like 1
Link to comment
  • 0

I can ask ChatGPT to clean up my notes after I take them in Evernote, but Evernote has still decided to add "AI Cleanup" to the application, removing the need for me to utilize two different applications. I am not sure "free alternative" availability should trump the prioritization of feature enhancement requests.

Link to comment
  • 0
  • Level 5*
3 hours ago, Paul Olson said:

I don't see why this would not be relevant to others who use Evernote as a document storage solution

I've been a user for 15 years,  have 62,000 notes up to 100MB each and I've only ever used OCR on an image once that I can remember - and that was because I saw it in Google images and thought "cool!" so tried it out.  Once.  

Ad if you're trying to make the argument that you didn't need AI Cleanup because it's already available from ChatGPT,  doesn't exactly the same argument apply to image OCR which is available from Google?

AFAIC I can file my images and search them for their text content if I need to.  I don't need to extract that text for use elsewhere...

Link to comment
  • 0

It's always worth remembering that EN OCR is not the same as traditional OCR. The one to many algorithm allows great searching of poor quality images,  with some false positives, but does not easily translate into  normal text. 

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...