Contents x
    Creating an Optical Character Recognition Application (OCR)
    • 08 Feb 2024
    • 3 Minutes to read
    • Print
    • Dark
      Light
    Contents

    Creating an Optical Character Recognition Application (OCR)

    • Updated on 08 Feb 2024
    • 3 Minutes to read
    • Print
    • Dark
      Light
    Article summary

    34 STEPS

    1. Select Character Recognition from the Application Builder. You may notice many different options, but for now we will select Character Recognition under Vision AI Applications.

    Step 1 image

    2. Select the Name text field to insert a name of the OCR (Optical Character Recognition) Application.

    Step 2 image

    3. Type the Name of the OCR application. When naming the application, it is important to name it something meaningful that directly relates to the application purpose. In the current scenario we named it 'automobile-plates'

    Step 3 image

    4. Click Next blue button, and you will proceed to the Input Subjects page.

    Step 4 image

    5. Click Next blue button, and you will proceed to the Input Subjects page.

    Step 5 image

    6.

    Step 6 image

    7. Type an Input Subjects name. Again please consider a name that will help you differentiate the Input Subject in the future. We have chosen 'automobiles' for the current scenario.

    Step 7 image

    8. Click Create a new subject in the current scenario, named as 'automobiles' 

    Step 8 image

    9. Click Next blue button in the end of the page, so you can proceed to the Output Subjects page

    Step 9 image

    10. The Output Subject of the OCR application is where you will find the labeled and sorted images. Usually the Output Subject gets populated with media (images/ videos) after successful Training of the Application.

    Step 10 image

    11. Type Output Subjects name. Again please consider the name, as it should be directly related with the Subject Of Interest displayed in the image. We have chosen 'automobiles' for the current scenario.

    Step 11 image

    12. Click Create a new subject in the current case scenario, named as automobiles

    Step 12 image

    13. Click Next

    Step 13 image

    14. Click Character Set. Here you can add more characters which could be detected in the images. You can also add Donated Subjects.

    Step 14 image

    15. Scroll down and click Complete

    Step 15 image

    16. Select the Input Subject which is always displayed in the beginning of the Application Pipelines. In the current scenario select automobiles.

    Step 16 image

    17. Click Upload Media to add the initial Dataset to to the Input Subjects. Once it has been uploaded to the Input Subject, the Media goes for training by the User.

    Step 17 image

    18. Click select files or select folder to upload the initial Dataset of images/ videos to the Input Subjects of the Application

    Step 18 image

    19. Click Next

    Step 19 image

    20. Click Upload

    Step 20 image

    21. Click Replays

    Step 21 image

    22. Click automobiles

    Step 22 image

    23. Click highlight

    Step 23 image

    24. With your mouse, click and drag highlight, then drop it on highlight

    Step 24 image

    24b. Drop

    Step 24b image

    25. Click highlight

    Step 25 image

    26. With your mouse, click and drag highlight, then drop it on highlight

    Step 26 image

    26b. Drop

    Step 26b image

    27. Scroll down and click Force Feedback

    Step 27 image

    28. Click Replay Subjects

    Step 28 image

    29. Click No Feedback

    Step 29 image

    30. Select the text field

    Step 30 image

    31. Type the presented characters in the image

    Step 31 image

    32. Select Confirm blue button to Submit the Feedback for the image.

    Step 32 image

    33. During providing feedback there are also: Back, Sideline and Skip options.

    Step 33 image

    34. That's it. You're done.

    Step 34 image

    Here's an interactive tutorial

    ** Best experienced in Full Screen (click the icon in the top right corner before you begin) **

    https://www.iorad.com/player/2274890/Cogniac---How-to-create-an-OCR-app-and-train-it

    I. In which cases we should use / what the app is doing

    Optical character recognition (OCR) is a computer vision technique that can extract text from images and videos. The OCR app identifies the individual characters in an image and maps them to their corresponding characters in a known alphabet.

    II. Use case / where the app can be used in real life:

    • It can be used for speed cameras to detect car plates and convert visualized characters.

    • Digitizing paper-based documents - The OCR app can convert scanned documents, such as invoices, contracts, and medical records, into editable and searchable digital formats, saving time and improving efficiency for businesses and organizations.

    • Automating document processing - the OCR app can automate document processing, such as extracting data from forms or receipts. It can help reduce errors and improve the accuracy and speed of data entry.

    III. What could cause issues while using the app:

    • OCR apps are highly reliant on the quality of the image being processed. If the image is blurry or noisy, the OCR app may struggle to identify the individual characters accurately. This can lead to errors in the output generated by the app.

    • Capturing media from various angles can cause significant issues for the OCR app when recognizing characters accurately. To ensure that the OCR app functions optimally, capturing media from the same angle is essential.

    • OCR struggles with handwritten documents, especially if the writing is unclear.

    Was this article helpful?
    What's Next