Google OCR incorrect text detection - google-cloud-vision

I'm using Google AI to get gext from image. It is using DOCUMENT_TEXT_DETECTTION,
but the data is incorrect.
Do you have any suggestion?
[Update]
expected result should be:

I would recommend for you to either:
Submit a Public report on Issue Trackers since OCR issues are better supported there. (Recommended)
Update your question with your request process; JSON, Client Library or API as
in Detect text in images.
Note that Supported Images are subject to these File formats, Image sizing and File size.

Related

PDF OCR Google cloud vision keep spacing

I don't know if it's possible, but I'm receiving forms in PDF format. I need to take the text from the PDF in the exact same position it is in the file and place it in a text document I can parse like:
Line 4 startCharacter 50 endcharacter 60
This would give me whatever text is in that position.
Is this possible?
At the moment, it is not possible to do this. I found a Feature Request made to Cloud Vision API to take a PDF file and export it as a searchable PDF which might resolve this issue. I recommend you to subscribe to the Feature Request (click on the star next to the title) so it can get more visibility.
In the meantime, you can check the documentation on PDFs recognition to try it out and see if you can get the desired behavior.
If the forms you are working with have a determined format, you might be able to solve the issue by going through the TextAnnotation response from the API. The response gives you the text, plus additional insights on it, like the pages, paragraphs, etc.

How to share only an image via Linkedin API

Using Linkedin via Buffer, I can share content with only an attached image. ie no accompanying comment, url, or any text. It is also possible via the web interface.
However, when using their share API, you are forced to include a URL and it displays with the content. Buffer must be using the API in some way, so how do you get around this?
I have been looking for the same solution.
How to get large LinkedIn Image Share Format
basically you just pass the image url in submitted-url field and don't pass submitted-image-url paramter in the json.
I was unable to find a this solution in google and in stackoverflow. I hope this help someone in the future.

Reading a PDF back from an iFrame?

I have a PDF document that is getting generated on the fly, and rendered on the fly to an iFrame within a radwindow. Basically the document is already largely prepopulated, however the user will still have a chunk of information that they are required to enter. I've found a good amount of information about sending a pdf TO an iframe, but not much information about going the other way. I have a button within the radwindow that can access the iframe object, however I'm somewhat lost as to where to go from there.
EDIT: The PDF is an editable form. I'm trying to pull back the entire PDF document as is, after the client side makes their entries to the form.
I think you'll need to send the file to the user so they can edit it locally and instruct them to upload it.
The content-disposition header with value attachment can help with the first task and you can use RadAsyncUpload to upload it: http://demos.telerik.com/aspnet-ajax/asyncupload/examples/overview/defaultcs.aspx.
I am not aware of ways to tap into the PDF viewer plugin the browsers use to show the PDF. Perhaps there is API from Adobe or some other third party plugin but that would rely on them and is out of your control.
Perhaps the JS PDF viewer from FireFox has something: https://mozillalabs.com/en-US/pdfjs/ but I don't know how stable and usable it is.
As per what was described in the comments, I ended up using postbacks through the PDF's themselves along with 1 pixel fields to store data required to identify the documents. It's a little hacky, but functional. I'm leaving this as an actual answer as this is as close to a real solution to the problem I originally had. This has been up and running for close to 4 years in this manner, and thus far hasn't caused any issues.

Why does Google Analytic request a GIF file?

Why does Google Analytic request a GIF file?
Is it because the GIF allows access to more data than JavaScript alone. Is it to get the IP address of the user?
Google's javascript has to transmit the details of your page view to their servers some how. Ajax cannot be used across domains, so the only way to submit the information is to request a file from a Google server, passing the necessary information in the query string. To facilitate this, some form of content must be requested using a standard HTML tag. The easiest way to do this without impacting the page itself is to load a 1x1 transparent gif. Note that in the case of the Google script (and others), this image isn't actually added to the page. It's merely loaded via a javascript statement
var img1 = new Image();
img1.src = 'http://path/to/file.gif?otherinfohere';
This loads the image without adding it to the page. The information could also be loaded using a script tag like so:
<script src="http://path/to/script.js?otherinfohere" type="text/javascript"><script>
However, users are more likely to have javascript blocked than images, so the safer route is to request an image. As to why they would use a gif instead of say, a jpg, a gif is safer in case a rogue browser actually adds the image to the page. The transparent gif is unlikely to negatively impact the layout, where as a 1x1 jpg would leave a 1 pixel dot somewhere on the page.
Edit: To add to my comment about users having blocked javascript, a gif request containing static information can be added inside a noscript tag to allow basic tracking even in the event that javascript is disabled. To my knowledge, GA doesn't do this, but some other web analytics providers do.
Even with JavaScript enabled, analytics requests a GIF file. If you look at the GET params of the image, it contains a lot of information about the browser. Stuff like utmsr=1280x1024 (the screen size). Google Code has a list of parameters.
It uses the image request to send information about the browser without an XMLHttpRequest.
Now, to actually answer the original question, Google is probably doing this to get around cross-domain XMLHttpRequest restrictions.
http://www.perlmonks.org/?node_id=7974
The smallest transparent GIF is 43 bytes.
http://garethrees.org/2007/11/14/pngcrush/
The smallest transparent PNG-24 (which can't be shown by older browsers too) is 67 bytes.
http://www.techsupportteam.org/forum/digital-imaging-photography/1892-worlds-smallest-valid-jpeg.html
The smallest (opaque) JPEG is 134 bytes.
The math is simple! Bigger size = more costs.
you can use the __utm.gif tracker without javascript (w some server help)
you can use it in an email message (w some programmatic help before sending the email)
Urchin was developed before AJAX was popular (2005)
It has nothing to do w cross-domain. They could have used JSONP for that.

Capture screenshot of website on the client (Javascript or flash)

Is there some method to issue a screen capture(browser window content only) from the browser with javascript or a embedded flash object etc so that a full quality image of the page content be saved or printed or an alternative approach.
I have a web app (asp.net 3.5) with google maps and other ajax operations client side like a custom tile server. I have been trying to implement a way for the user to print good quality captures of the webpage.
I have used the basic Window.Print() but in both IE and FF there many artifacts within the google maps and some items such as the popped up bubble doesnt print. I have experimented with save pdf thru cutepdf(just to post an example here) and the quality thru window.print() is low too.
For example, A screenshot with FireShot addon is perfect and what I want the client to have. however that is FF only and I cant ask the clients to install addons/activex controls on their browsers.
Have a look at this download example zip file(4mb) with:
Example screen shot using FireShot
(example of what I want to achieve
thru a html/JS button with in the
page)
Firefox Window.Print() result
(cutepdf used to save as pdf)
IE Window.Print() result (cutepdf
used to save as pdf)
note in 2,3 the little bubble is not printed even when open.
For now, I have added the function on my site to go fullscreen and guide the user to take a screenshot or call the window.print() function.
I am still looking for a method to print/capture my page.
are there any flash/activex controls that I can include in my page and thru them provide a quality print mechanism?
Thanks again for all the help but I still need more. :)
Thank you in advance.
http://rapidshare.com/files/311849636/Print_examples.zip.html
You'll go to all that work only to find that a simple app like Snagit will do the job. Building a SnagIt Screen Capture Plugin
The only way to reliably provide a high-quality print version of whats on-screen in a rich web application is to use the client-side, say JavaScript, to send the server precise information about the current state (where bubbles are, etc.) and use that to generate an image that mimics the positioning. Convert that image to a PDF or what-have-you, then send to the client for download.
Google also has a Static Maps API that might give you good results. I looked into it myself once, and only didn't go with it since (at the time) there were limitations on how many points they could support in a polyline.
I don't think this is possible. It would be quite a security risk to be able to capture the user's screen through scripting (imagine bad sites capturing screen information).
No there is not, though it would come in quite handy at times for bug reporting etcetera.
You will probably get the best result by creating a separate version of the page as a PDF have that being generated. It's no quick fix by all means, but you'll get superb print quality and total control over everything. The map part will probably be a bit tricky though as you need to get the map as a bitmap on the server somehow, and if it's not in flash on the client I don't know how you'd do that.

Resources