It is a smart labeling model that performs OCR (Optical Character Recognition) tasks that can recognize diverse languages.
It is available when the task type is set to
Polygon in the project's
- Create a project by setting the "Task type" to
"Polygon"in the project's
- After selecting the corresponding class, and then add
"Attributes by each class".
- After entering the Attribute name that you want to use, set "Attribute entering method" to
- Put "entering type" as "characters", "Minimum number of entering characters" as 1, and "Maximum number of entering characters" as 1,000.
- Click the
"completion"button at the bottom right.
- After selecting the folder or file that you want to perform OCR-Smart Labeling in the dataset of the project and then click "Smart labeling".
"Select a task AI model", select the pre-set model that corresponds to the language that you want.
- OCR: Korean/English/Numeric
- OCR (Japanese): Japanese/English
- OCR (Russian): Russian
- 'Set the class for that you set
"Attribute"at STEP-1 to match the "text" class of the Pre-set Model.
"setting others", set the "Threshold" standard value to the minimum value of "0.2".
- By clicking the
"Completion"button at the bottom right, proceed with OCR Smart Labeling.
- A square box in the form of the "Polygon" will be created in the area that has letters.
- Within the generated "Polygon" area, the value which corresponds to the "Attribute Name" designated at the project's "Labeling Settings" is characters inferred by the OCR model will be entered automatically.
- Workers will inspect or work again on the results of the OCR model inference.
- Information about images
- Resolution: HD or higher
- Minimum size: If the height of the text is less than 8px, the performance of the OCR model will be degraded.
- For OCR models, Custom Model learning is not supported.
If you have any other inquiries, please get in touch with us at [email protected]
Updated 9 months ago