Building an AI-Based OCR Solution in Oracle APEX Using Google Vision API

Introduction

While working on one of my Oracle APEX applications, I encountered a common business problem — users were manually entering information from uploaded images such as certificates, labels, and printed documents into the system.

This process was slow, repetitive, and error-prone, especially when users handled multiple records daily. I wanted to automate the text extraction process directly inside Oracle APEX without introducing complex third-party desktop software into the workflow.

After exploring different OCR options, I decided to integrate Google Vision API with Oracle APEX using PL/SQL and REST APIs. The goal was simple:

• Upload an image
• Extract text using AI OCR
• Display the extracted result instantly inside the application

Application Flow

The complete OCR workflow looks like this:

Creating the Upload Interface in Oracle APEX

The first step was building a simple upload interface where users could upload images and trigger OCR processing.

1. File Upload Item

For image uploads, I used the built-in File Browse item available in Oracle APEX.

Page Item

P2_IMAGE_UPLOAD

This allowed users to upload:

• JPG
• PNG
• JPEG

files directly from desktop and mobile devices.

One thing I liked about using the native APEX upload component is that uploaded files are automatically stored temporarily inside:

APEX_APPLICATION_TEMP_FILES

which made the backend processing straightforward.

2. Extract Text Button

Next, I created a button to trigger the OCR process.

Button Name

EXTRACT_TEXT

When the button is clicked:

• The uploaded image is processed
• A PL/SQL process starts
• The REST API request is sent to Google Vision API
• OCR extraction begins

Initially, I tested the process using Dynamic Actions, but later shifted most of the logic into a PL/SQL process to simplify debugging and response handling.

3. OCR Result Display Field

To display the extracted content, I used a Text Area item.

Page Item

P2_DISPLAY_TEXT

Once the OCR process completes, the extracted text is automatically displayed inside this field.

Configuring Google Vision API

Before integrating with Oracle APEX, Google Vision API needs to be configured.

I created a Google Cloud project and enabled the Vision API service.

The setup process was fairly simple.

Step 1 — Create Google Cloud Project

Inside Google Cloud Console:

• Create a new project

Step 2 — Enable Vision API

Navigate to:

APIs & Services → Library

Then enable:

Cloud Vision API

Step 3 — Generate API Key

Under:

Credentials → Create Credentials

generate an API key that will later be used inside the Oracle APEX REST call.

Reading Uploaded Files in Oracle APEX

Once the image is uploaded, Oracle APEX stores the file temporarily inside:

APEX_APPLICATION_TEMP_FILES

The first step in the PL/SQL process was reading the uploaded BLOB file.

SELECT blob_content
INTO l_blob
FROM apex_application_temp_files
WHERE name = :P2_IMAGE_UPLOAD;

This part was straightforward, but during testing I noticed that large image uploads significantly increased processing time.

Converting Image to Base64

Google Vision API accepts images in Base64 format.

So the uploaded BLOB needed to be converted before sending the API request.

l_base64 :=
    replace(
        replace(
            apex_web_service.blob2clobbase64(l_blob),
            chr(10),
            ''
        ),
        chr(13),
        ''
    );

One issue I faced here was line breaks being automatically inserted into the Base64 output. Removing carriage returns and newline characters was necessary to avoid malformed JSON payloads.

This small cleanup step solved several API request failures during testing.

Building the JSON Request Payload

After converting the image, the next step was constructing the JSON payload expected by Google Vision API.

l_request_body :=
'{
  "requests":[
    {
      "image":{
        "content":"' || l_base64 || '"
      },
      "features":[
        {
          "type":"TEXT_DETECTION"
        }
      ]
    }
  ]
}';

I initially tested multiple OCR detection types, but for my use case, TEXT_DETECTION provided the best balance between speed and accuracy.

Sending REST API Request from Oracle APEX

The REST request was handled using the APEX_WEB_SERVICE package.

l_response := apex_web_service.make_rest_request(
    p_url         => 'https://vision.googleapis.com/v1/images:annotate?key=YOUR_API_KEY',
    p_http_method => 'POST',
    p_body        => l_request_body
);

This was one of the most interesting parts of the implementation because Oracle APEX handled the external REST integration smoothly without requiring additional middleware.

During testing, I also added proper exception handling to capture API failures, invalid images, and timeout scenarios.

Parsing OCR Response

Google Vision API returns a nested JSON response containing the detected text.

To extract the OCR content, I used JSON_VALUE.

SELECT json_value(
    l_response,
    '$.responses[0].fullTextAnnotation.text'
)
INTO l_text
FROM dual;

The extracted text was then assigned directly to the page item.

:P2_DISPLAY_TEXT := l_text;

At this stage, the OCR result immediately appeared on the screen after processing.

Final Result

The final solution significantly reduced manual data entry effort inside the application.

Users could now:

• Upload an image
• Click a button
• Extract text instantly using AI OCR

directly within Oracle APEX.

The implementation also demonstrated how easily Oracle APEX can integrate with modern AI services using REST APIs and PL/SQL.

Conclusion

This implementation started as an experiment to reduce manual typing effort in an Oracle APEX application, but it quickly became a practical AI-powered feature with real business value.

What I found most interesting during this project was how effectively Oracle APEX handled:

• File uploads
• REST integrations
• JSON parsing
• Dynamic UI updates

without requiring complex external frameworks.

For developers exploring AI integrations in Oracle APEX, OCR is an excellent starting point because it combines practical business usage with modern AI capabilities.

The same approach can later be extended for:

• Intelligent document processing
• Invoice automation
• Identity verification
• Barcode scanning
• AI-assisted data entry
• Smart validation workflows

As AI services become more accessible through REST APIs, integrating intelligent automation into Oracle APEX applications is becoming much more achievable than before.