DetectAreasMode
Contents
[
Hide
]DetectAreasMode enumeration
Determines the type of neural network used for areas detection.
Members
| Member name | Description |
|---|---|
| NONE | Doesn’t detect paragraphs. Better for a simple one-column document without pictures. |
| DOCUMENT | Detects paragraphs uses NN model for documents. Better for multicolumn document, document with pictures or with other not text objects. |
| PHOTO | Detects paragraphs uses NN model for photos. Better for image with a lot of pictures and other not text objects. |
| COMBINE | Detects paragraphs with text and then uses other NN model to detect areas inside of paragraphs. Better for images with complex structure. |
| TABLE | Detects tabular structures in the image and extracts text from individual cells. Recommended for scanned spreadsheets, reports, and other table-based documents. |
| CURVED_TEXT | Automatically straightens curved lines of text in the image, improving recognition accuracy and allowing more text to be recovered and extracted. Requires significant processing power and RAM. |
| TEXT_IN_WILD | A super-powerful neural network specialized in extracting words from low-quality images such as street photos, license plates, passport photos, meter photos, and photos with noisy backgrounds. |
| LEAN | Prioritizes speed and reduces resource consumption by omitting support for complex layouts. Suitable only for simple images with a few lines of text without illustrations or formatting. |
| MULTICOLUMN | Detects large blocks of text formatted in columns. The best choice for multi-column layouts such as book pages, articles, or contracts. |
| UNIVERSAL | Detects all blocks of text in the image, including sparse and irregular text on photos. A versatile option for most images, except for tables and multi-column layouts. |
See Also
- namespace aspose.ocr
- assembly Aspose.OCR