Using CLOVA OCR
  • PDF

Using CLOVA OCR

  • PDF

You can extract text from images and documents in a quick and easy manner with CLOVA OCR. Follow the instructions below to use the OCR builder.

How to use CLOVA OCR

Connect to the console and follow the steps below to easily use the service.

  • Step 1. Request subscription to CLOVA OCR and agree to Terms and Conditions
  • Step 2. Create domain
  • Step 3. Create template
  • Step 4. Test and analysis
  • Step 5. Components
  • Step 6. Settings
  • Step 7. Manage deployments

Step 1. Request subscription to CLOVA OCR and agree to Terms and Conditions

To use CLOVA OCR, you must first request subscription to the service and agree to the Terms and Conditions.

The Terms and Conditions of this service explain the terms regarding the retention and use of data generated in CLOVA OCR, consignment of personal information, and obligations of the company and customers. Please make sure to check the details of the Terms and Conditions before planning your use of the service.

Go to CLOVA OCR Terms and Conditions

Click Request subscription and agree to the Terms and Conditions to move to the next step.

Step 2. Create domain

Create a domain to begin producing your CLOVA OCR.

A domain is a standard unit for the CLOVA OCR service.
For example, when you create a "written notice" OCR, the domain is the unit that includes all recognized notice templates. In this way, the creation of templates, linkage of channels, and statistical information are operated and managed based on a single domain.

① Click Create domain to see the menu to create a General, Template, or Document domain.

  • General/Template domain: Text OCR to extract only text. A template builder is supported through which you can extract recognized values by directly specifying an analysis area, test them, and send the results.
  • Document domain: The key values are automatically created since it is mounted with a specialized model engine, built on machine learning, and understands a document's semantic structure.

Create General/Template domain: Enter a domain name and domain code. (Document OCR will be explained with in detail at the bottom of the domain creation guide, "add specialized model" section.)

  • Domain name/code: Unique value between 1 and 50 characters that isn't a duplicate.

  • Characters available for input in the "Input field" of the OCR builder: Korean, English (half-width uppercase/lowercase letters), Japanese (full-width katakana/hiragana), Chinese, and some special characters (? & 。_ - × 々).

    However, only uppercase and lowercase English letters (half-width uppercase/lowercase letters) and some special characters (. ? & 。_ - × 々) are available for input for domain codes.

  • Supported languages: Select a language between Korean or Japanese. As English is supported by default, the letters of both the chosen supported language and English are recognized for documents that contain English.

  • Service types: General type, which can only extract text, and Template type, which provides a template builder, are supported. The General type only allows the creation of a single General domain (Text OCR execution) per region. Selection of the Template type displays recognition models and service plan items.

ocr3_1_en(3).png

  • Recognition model: Divided into the Basic model, which provides basic recognition features, and Premium model (including handwriting recognition), which provides a further enhanced recognition rate. Template layouts are supported by all service models. Value type settings for multi-box and field input values are features exclusive to the Premium model.
Recognition model Basic Premium Description
Model recognition rate Typeface and handwriting recognition Typeface and handwriting recognition Provides high-performance OCR recognition model optimized for usage in major businesses
Appropriate documents Certificates, documents with a fixed form Handwritten sign-up application forms, financial documents, etc.
Recognition template layout Provided Provided Sets recognition area
Multi-box Not provided Provided Combines single box templates to control number recognition area, etc.
Checkbox Not provided Provided Provides recognition in the form of a checkbox
Field type Not provided Provided Sets the type to recognize recognition values only as numbers
  • Service plan: CLOVA OCR applies charges based on the service plan for each domain. Basic maintenance costs are incurred for all service plans, except for free services in the Basic recognition model, even if the OCR API isn't called. The number of OCR API calls provided varies by service plan, and additional charges are incurred if the number of calls exceeds your plan's limit. Click "Pricing Information by Service Plan" to check pricing plan details.

    1205_ocr_1_en.png

  • Notes for creating domains
    • CLOVA OCR applies charges based on the service plan for each domain.
    • Basic maintenance costs are incurred for all service plans, except for free services in the Basic recognition model, even if the OCR API isn't called.
    • The number of OCR API calls provided varies by service plan, and additional charges are incurred if the number of calls exceeds your plan's limit. Pricing Information by Service Plan

③ You can check the domain list after the domain is created. General domains, which can only extract text, display the Text OCR button (API Gateway linkage) and the Demo button under Action menu, and Template domains display the Run template builder button.

④ Click the Template builder button under the Action menu of the selected domain to check domain information and run the OCR builder in a new window.

ocr6_1_en(3).png

⑤ Click the Text OCR button under the Action menu. A pop-up window for setting linkage with General OCR, which can extract the entire text without making template settings, appears. Automatic interlock is enabled through custom API Gateway settings.

ocr7_1_en(3).png

  • Automatic interlock through custom API Gateway settings
    • Click the Text OCR button to display a window for setting linkage with external addresses.
    • OCR Invoke URL refers to the OCR API address to be entered in the endpoint of external linkage.
    • If a subscription request has been made for API Gateway, click the Automatic interlock button for easy automatic linkage.
    • Up to 10 pages is supported for PDF Recognition. A charge of one General OCR call is incurred for each page of the PDF document. General OCR Pricing information

⑥ Click the Demo button under the Action menu. This provides the Text OCR results of an uploaded image in lines, without linkage with API Gateway, so that you can visually check the data and download them.

OCR_EN_1-1

OCR_EN_1-2

The table extraction feature has been added

⑦ When enabled toggle, it automatically recognizes the table area in the document and reproduces it in a structured format with the text.
OCR_table_en

Document OCR, which has a specialized model optimized for different document through AI learning in advance mounted, has been added.

You can request subscription to Document OCR in the NAVER Cloud portal. Please contact Support for consultation related to any specialized models.

Please accept the Agreement for the technical and administrative levels of CLOVA OCR (specialized model) to move on to the next step.

① Click Specialized model settings to select the specialized model you are going to use, and complete the settings application.

② Click Create domain to create a Document domain.

③ Enter the domain name and domain code.

  • Recognition model
    • Business registration certificate (KR): Recognizes (Individual/Corporate) business registration certificate, proof of business registration, mail order distributor registration certificate, business declaration certificate, and certificate of registered seal
    • Credit card (global): Extracts card numbers and expiration date, supports recognition of a vertical format credit card
    • Receipt (KR): Extracts key information such as store information, payment details, payment method, amount, etc.
    • ID (KR): Recognizes resident registration certificate, driver's license, and passport
    • Business card (KR/EN/JP): Extracts information such as name, company name, department, phone number, address, email, etc.

④ You can view the list of domains after domain creation is completed. Click the Document button in the Action menu to display the image recognition request guide and API Gateway settings pop-up window.

A copy feature for domains has been added.

Click the Copy domain button to create a new domain that maintains the existing settings. You can click the checkbox on the left side of the service type Template or Document domain, copy the domain, and change the recognition model (General type not supported). When changing the recognition model from Premium to Basic after copying a Template domain, features exclusive to the Premium service model, such as Value type settings for multi-box and field input values, aren't copied. Only the features for the Basic service model are supported.

A search feature for domains has been added.

The feature for searching domain names and domain codes is provided on the domain list screen. Use Filter to view domains by supported languages, recognition models, and service plans. For General domains of Text OCR, you can select Filter > Recognition model > Premium to apply General in service plans.

Step 3. Create template

OCR builder is a web console that allows you to create and test templates for extracting text from image documents. It provides various features required by CLOVA OCR.

Click Template list in the left menu.

① Click Create template to open the page for template creation where you can enter the template's basic information.

  • Register the template name under Template name and click Confirm.
  • Enter the template's basic information to enable the areas of Main sample name and Specify analysis field in Set main sample below.

ocr9_1_en(3).png

② Upload an image sample to Set main sample.

  • Recommended uploads are image files whose size is under 20 MB. JPG/PNG/PDF/TIFF are supported. Images should be above A4 150 DPI, and have a resolution between 10 px and 1960 px, based on the major axis.
  • You can specify the analysis field after uploading the main sample image.
  • High-quality images of 150 DPI or higher, with little distortion and damage, are recommended for images entered via APIs in the future.

ocr10_1_en(3).png

③ Specify the title area of document.

  • Upload a main sample image and drag the rectangle for specifying the title area of the document.
  • Clicking an already specified area enables the rectangle for specifying the area, and clicking X cancels the area specification.
  • Clicking an already specified area enables the rectangle for specifying the area, and dragging the dotted line with the cursor resets the area. Click V to save the reset area.

ocr17_1_en(3).png

④ Enter the main sample name.

  • The main sample name entered is used for the document's template classification.

ocr18_1_en(3).png

  • The Confirm button is disabled until the main sample name is entered. Enter the main sample name to enable the Confirm button. When you click Confirm, notification windows for saving and synonym registration will be displayed.
  • You can freely modify the main sample name in the corresponding field.
  • Register words that are recognized in a similar manner to the main sample name in Manage synonyms to enhance classification performance. When entering multiple synonyms, distinguish them by using a comma (,) in between synonyms (e.g., local taxation notice, local tax, local tax notice).
  • The main sample name of Manage synonyms can't be changed. You can only register and modify synonyms.

ocr19_1_en(3).png

⑤ Specify an analysis field.

Click + Add field in the Specify field tab to specify an analysis field. Specifying multi-box and checkbox is a feature exclusive to the Premium recognition model and is disabled in the Basic recognition model. More information about the detailed features of multi-box and checkbox will be provided in the field specification guide for the Premium recognition model below.

  • Specify field

    ocr25_1_en(3).png

    • You can set the scope of analysis for each field. Click the + Add field button to display the rectangle for specifying a new field. Adjust the size of the rectangle and specify the analysis area.

    • Clicking an already specified area enables the rectangle for specifying the area, and clicking X cancels the area specification.

    • Clicking an already specified area enables the rectangle for specifying the area, and dragging the dotted line with the cursor resets the area. Click V to save the reset area.

    • Click the Delete icon on the right side of the field box to delete a specified field (Delete Field 01).

      ocr25_2_en(3).png

    • Specify a field area and enter the name of the field.

    • You can freely modify the field name in the corresponding field.

    • Click the Settings icon on the right side of the field name to open the field setting windows for Value type and Register terms.

    • Value type: You can specify to recognize a combination of text and numbers by selecting All, and only numbers by selecting Numeric. Use the Value type feature to specify the form for values to be entered in the field in advance so that only specific values will be extracted.

    ocr23_2_en(3).png

    • Register terms: You can register a glossary containing domain-specific words and synonyms. When a dictionary is being registered, only the representative terms included in the dictionary are recognized. Synonyms, upon identification, are substituted by representative terms and displayed. It is processed as a failure when words other than the representative terms and synonyms are recognized.
  • Specify field in the Premium recognition model

    • Multi-box: If input boxes are separated, as in bank account numbers or mobile phone numbers, you can specify a multi-box to recognize and extract characters inside individual boxes.
      Set the field outside the frame of the multi-box so that only the characters inside the box you wish to extract are recognized. (Multi-box frame not recognized/Characters inside 12 multi-boxes per field recognized)
    • Checkbox: Value of True is returned if unstructured characters like V, X, O, and · are recognized in the rectangular (ㅁ) or circular (O) check mark box, and value of False is returned if not recognized. Check marks can be recognized with OCR in various kinds of documents such as surveys and product sign-up documents.
    • When specifying a checkbox, the field must be set so that the border of the input box is visible. Try to match the multi-box border and field domain box together as much as possible.
    • The specification should only allow the checkbox to be recognized within the specified area without other characters or shapes. Check marks might not be recognized correctly in the following cases: they are outside the checkbox area, they are blurry or erased, they are duplicated in the same checkbox due to modification, or "check marks and guide characters" have already been entered in the checkbox for the customer to check off.

ocr23_3_en(3).png

  • Change main sample image

    ocr27_1_en(4).png

    • Click the Change main sample image button to change the registered main sample and modify its details.

    • Click the Confirm button in the pop-up window for modifying the main sample to move to the page for changing the main sample image.

    • The template's basic information, such as the template name and usage model registered at the time of template creation, are entered automatically.

    • Settings information for the main sample, such as the main sample image, main sample name, and analysis field specification, are displayed as blank.

      ocr28_1_en(4).png

⑥ Click the Save button to complete template creation, and then click Template list on the left menu to view the created template.

ocr29_1_en(3).png

  • Click Modify main sample to move to the page for modifying the image and details.

  • Select the checkbox on the left of the template ID from the template list, and then click the enabled Delete button to delete the created template.

    ocr31_1_en(3).png

  • If you own a verification server via the verification feature, then you can have it linked and proceed with the verification of each template.

    ocr32_1_en(3).png

⑦ Click Sample list on the left menu.

1205_ocr_3_en.png

  • View sample: Click individual sample numbers on the sample list to view the uploaded main sample image.

    ocr36_1_en(3).png

Step 4. Test and analysis

The analysis test and download feature is provided after the image file, from which the text is to be extracted, is uploaded.
You can check the number of daily API calls for a service plan and the results of external verification by template on the integrated indicator page.

Click Test on the left menu, then select the test condition for "Beta/Template analysis."

  • Beta test: You can test the automatic template classification and analysis area in the same environment as service deployment. Beta testing and service deployment can be performed after beta deployment.
  • Template analysis test: Test results for templates and models with set analysis areas are provided.

ocr37_1_en(3).png

Click the Find file button to display a pop-up window for uploading image files. Select a template suitable for the test image → Upload image → Click the Test button to check the analysis results.

ocr38_1_en(3).png

  • 300 free tests, which is the total number of beta and template analysis tests, are provided each month.

    ocr39_1_en(3).png

  • Test results can be downloaded in CSV and JSON formats, and analysis results are provided in a table on the right side of the test results.

    ocr43_2_en(3).png

Click Integrated indicators on the left menu.

  • A period of up to 90 days can be selected.

  • You can check the indicators for each selected template. The selection of all templates displays indicators for all domains.

  • Download integrated indicators: The selected indicators can be downloaded as Excel files.

  • Number of API calls: Refers to the number of APIs called during the set period of the service plan in use.

  • Verification results: You can check the external verification results by template and the number of verifications for corresponding domains. When you mouse over a verification result graph, the numbers of VALID, UNCHECKED, ERROR, and INVALID are displayed.

  • Number of template recognition failures: When there are no template matches and no recognition results, it is counted as a "template recognition failure," and the number of failures during the set period is displayed.

    1205_ocr_4_en.png

Step 5. Components

Glossary

  • If you specify specific fields to only recognize words from the dictionary, or if registered synonyms are recognized, then the results are substituted by representative terms and displayed.
    • Glossaries can be uploaded and downloaded in bulk as Excel files.
    • When you search for dictionary names, the search result of the corresponding glossary is shown. Clicking the Refresh button displays the first page of the glossary.

ocr47_1_en(3).png

  • Click the Create glossary button to open the "Create and modify dictionary" pop-up window.

  • Dictionary name: Enter a unique dictionary name that isn't a duplicate.

  • Representative terms: Enter the representative terms that make up a dictionary.
    Synonyms: Enter the synonyms of input representative terms. When entering multiple synonyms, distinguish them by using a comma (,) (e.g., local taxation notice, local tax, local tax notice).

    ocr48_1_en(3).png

Step 6. Settings

OCR builder provides easy setting features such as domain information, external linkage, and permissions management.

ocr51_1_en(3).png

External linkage

ocr50_1_en(3).png

You must set linkage with external platforms in order to deliver the final OCR detection and recognition values to the linked address.
Since the service should be provided securely to external users, it is linked with API Gateway for provision.

  • Automatic interlock through custom API Gateway settings
    • Click the Settings button to display the window for setting linkage with external addresses.
    • OCR Invoke URL refers to the OCR API address to be entered in the endpoint of external linkage.
    • If a subscription request has been made for API Gateway, click the Automatic interlock button for easy automatic linkage.

Verification linkage

ocr53_1_en(3).png

  • Transfer data
    • You can directly set and manage addresses to send OCR results.
    • Click the Create button and enter a name in the ${name to send results} format.
    • Enter a URL that begins with https:// in the URL field.

Manage permissions

ocr55_1_en(3).png

  • Add new admins and grant permissions.
    • Create URL for permission settings: Click the button to create an invite URL. You can grant permissions via the URL.

    • Click the Copy URL button to display the copy alert for Invite URL.

    • Click the Delete button to display the delete alert for Invite URL.

Step 7. Manage deployments

Beta deployment is the final storage feature reflected in the model. You can learn the samples, analysis areas, and glossaries that have been applied to templates. Beta testing and service deployment can be performed after beta deployment.

Click Manage deployments on the left menu to view the template list.

1205_ocr_5_en.png

  • Select template: Select a template from the drop-down box to display the corresponding template's deployment history.

  • Current deployment status: Displays the latest deployment status of the corresponding template.

  • Select a template to deploy from the template list to enable the Beta deployment button.

ocr60_1_en(3).png

  • Click the Beta deployment button to view the list of templates for beta deployment.

  • Click the Confirm button to proceed with beta deployment. Beta testing is available after completing the deployment. Information modified after testing the analysis result on the Test menu is applied through the new beta deployment.

  • Perform a service deployment of the latest beta deployment template to have completed the beta test.

ocr62_1_en(3).png

  • Click the Service deployment button located at the top of the OCR builder.
  • You can check the service deployment template list, domain name, and service plans in use in the service deployment window.
  • Click the Confirm button to proceed with service deployment.

1205_ocr_6_en.png

  • Deployment history by template is provided upon selecting individual templates in the drop-down box on the right side of deployment management.

ocr58_1_en(3).png

  • Selecting individual templates displays details about the deployment history such as the deployment time. You can check the history of beta, service, and beta deployment failures in the current deployment status.

Notes for using the service

  • The one-time call standard of Template OCR is the number of recognition areas of the template set in the builder (maximum of 50). Additional charges apply when this is exceeded. (For example, if the number of box areas of the Template is 130, then it is charged as 3 API calls.)

  • CLOVA OCR is committed to the provision of high recognition performance and fast service processing speed. Preparations are underway for features that will provide large recognition amounts in a short period or batch processing them in a given time. Currently, the recommended call performance per service account is 1 TPS at maximum. (Please contact Support for a higher call performance.)

  • If you request OCR recognition for a document rotated 45 degrees or more, then the recognition rate may decrease.


Was this article helpful?