OCR not working on Android-app scanned "documents"

[ronnyd 2]

 

I rely quite a bit on my Android Evernote app (currently version 10.12.1) to scan paper documents into Evernote.  To do so I use the option to scan it as a “document”, where there is edge detection and orientation correction. This works quite nicely most of the time though I admit I would like it to work even faster than it does.

My issue, however, is that the images generated using this process are not coming up in searches for words that are clearly visible on the documents that the Evernote app itself creates from these paper artifacts.

What I've been noticing is that if I create a note with the document scanned as a “document” from the app and additionally include the document as a photograph, then searching for words on the document returns the note in my search results, with the desired word highlighted only on the photographic representation of the document, not the actual “document.” This is extremely disappointing to say the least. I've even run at home quick test at home (in the browser version of Evernote, this shouldn’t be an issue, but just providing context) whereby the same document was used in three different notes, 1) with both the “document” and the photograph captures, 2) one with the photograph alone, and 3) one with the “document” alone. Unfortunately, but not unexpectedly, the search results yield only the versions with the photographic representation. 

Anyone have any ideas what's going on here? I'm about to file a support ticket / bug report.

[agsteele 3,056]

I'm assuming that you are on a paid subscription plan...

The first thing that the support team will request is that you upgrade to the most recent version of the app which is v10.28 (actually there is a version 10.29 slowly rolling out but 10.28 is the most generally available version).

OCR takes time to process since it is a server side process. So it isn't instant by any means.

[PinkElephant 8,738]

Tried it on iOS 10.29 (the client is not that important, since everything is created serverside).

No issue, search results show and are highlighted on scanned images as well. They were positively created by the build in scanner in document-color mode.

[ronnyd 2]

Thank you both for your replies. Yes, I am a premium personal subscriber.  I am delighted that it works for you, truly, but as I reported in my original post this is simply not the case for me. I believe that I've done my due diligence to ensure that any unreal expectations or insanity on my part are not the issue here.  I am on the latest version of the app that's available for my device. The Google Play store does not show an option to update the app.

I am aware that the OCR function is server side and so my device and app version should be irrelevant, as both OCR and searching are done on Evernote’s side. I wish this were a matter of me not being patient enough, but that does not seem to be the case. The most recently edited file DOES come back, but it’s one that photo only, not a scanned document. I simply did this test with the same receipt/document today to have something to share with this community and with support.

My befuddlement is that if the OCR has happened on the photograph why has it not also happened on the document scanned and in the same note?

As they say a picture is worth 1000 words so here's a few demonstrating when I'm talking about. I have the same receipt from the grocery store replicated across the three notes that I described above 1) one has both a photo and a scanned color document version 2) one has the scanned document alone and 3) one has the photo alone.  when I search for the word “juicy” looking for the line item on the receipt for the Juicy Fruit gum, I get only the notes in the search results that have the photo version. I don’t get the one that has only the scanned document version. Both the photos and the color scanned document were created with the Evernote Android app (latest version possible for me). Searches executed in Evernote Web Edition and in the Android app.  Same results (as expected.) What’s NOT expected is the lack of return for a scanned document.

[PinkElephant 8,738]

Just picked up the idea and scanned a receipt.

iOS app 10.29, iPhone 11, the in app scanner / camera. The modes I used:

• Scan, then document-color
• Scan, then foto
• Camera, then foto

All 3 from the same receipt, created within of a few minutes, synced directly with EN server. Only 4 minutes after creating the scans I searched for the word "Espresso", and got all 3 notes on top of the search list. The word was only contained in the pictures, EN had it OCRed nearly instantly.

When I look up the Picture Information, the two pictures saved as Foto are jpg-Files, the one taken as document-color as png-File. This means that choosing a different setting in fact has an influence. The word is highlighted in all 3 cases. Using the 10.33 client on my Mac brings up the 3 notes as well.

Conclusion: Maybe the Android app does it differently (you could check the file type created), but I have my doubts.

I would approach support - you can send them the link to this thread as well to show we had an intense discussion about it.

 

[agsteele 3,056]

I'm sorry that you continue with this issue. I'm not aware of any reasons why you would be limited to v10.12.1. It would be appropriate to raise a ticket. You're not getting the update that should reach you and the issue you've described continues.

One additional question, which version of the Android OS do you have?

[ronnyd 2]

@agsteele - Happy to share that information, but it shouldn't matter (not saying it doesn't matter, because we all know how software can be) since all of the issues I'm facing seem to be from a breakdown on server side processing (OCR and/or search). It's a Samsung Note 9, running Android 10, and Samsung One UI 2.5.

@PinkElephant - interestingly, I tried the same experiment, and the android app is producing jpg for all scans/photo captures: color documents, b&w documents, and photos scans. 

I tried my test again with a new document, explicitly saving as color and black and white, both in notes alone and alongside photos of the document.  Same results.  Search results yield only the notes containing the photo representations of the document with the searched word highlighted in the image, not the notes with the scans alone (B&W and color both).

I will be filing a support ticket and linking to this thread. Thank you both for your input.

[agsteele 3,056]

22 minutes ago, ronnyd said:

It's a Samsung Note 9, running Android 10, and Samsung One UI 2.5.

I asked, only to be sure what I said next was correct. You should be on Evernote 10.28 or 10.29 I don't know why the Play Store hasn't updated for you but your device should have the latest version. I wonder if Google Play has got stuck. Having the current version of the Evernote app could be the answer to your issue. I think you can clear the cache and memory for Google Play and Play Services to clean things up. You will, I think, need to log back in to Play to get going.

[ronnyd 2]

I'm actually really annoyed now. 😊 And thankful too.

I couldn't figure out how to get Play Store to allow the update so I uninstalled and reinstalled the app, forcing it so. In doing so I see that I'm now on version 10.29. So, something along the way was indeed stuck with the Play store updates.

What's really annoying though, and what doesn’t make sense, is now that I'm scanning documents with this updated version of the app the documents are being OCR’d appropriately and returned in search results. I am baffled by this. We've been told all along that OCR on a document happens on the server side so not sure what in the process of producing these scanned document iterations was broken previous to this version.  I'm still going to report this to the support team and I will inquire whether there's a possibility of putting my notes generated over the course of the past year back into the queue to be re analyzed for OCR. I haven't noted this before but I was and still am according to Google Play part of the beta tester program for Evernote. Not sure if that would have made a difference, but there it is.

[PinkElephant 8,738]

48 minutes ago, ronnyd said:

…. but I was and still am according to Google Play part of the beta tester program for Evernote. Not sure if that would have made a difference, but there it is.

Yes, it does make a large difference, and for me it is the most likely reason why your app was not updating. To make sure your app works on a regular basis, you need to exit the Beta program.

https://help.evernote.com/hc/en-us/articles/360033369173

Just read the first sentence: You locked yourself out ! After that, anything can happen as you experienced it, because the back office software on the server moves ahead, while you stay frozen on an older beta code level.

Did you tell support you were still in the beta when you contacted them ? They probably would have told you to exit beta first, then update and try again.

Which in the end you did yourself …

No idea about your old Scans now. Probably they won’t OCR by themselves even if now saved in a release environment. If not, try to export them out of EN, and bring them back in to newly created notes.

P.S. I never ever use a beta on any device in productive use - and I think I just got confirmed in that decision.

[agsteele 3,056]

Being on the beta could will account for you being stuck at 10.12

I think you have to exit the beta to reconnect with general release versions.

Sorry you're annoyed. I thought we'd made progress for you

[ronnyd 2]

I'm not annoyed at YOU, @agsteele! 🙂 Your tip got me to a better place!