Layout analysis
Business scenario description
The traditional character recognition ability can extract the characters on the picture and output the recognition results according to certain rules (such as from left to right and from top to bottom). However, when the picture appears a more complex arrangement structure, such as column, text mixed row, fixed rule output will cause unreadable results, affecting the next Process.
Layout analysis is a picture understanding atomic ability provided by UiBot Mage. The model will detect and analyze the content in the picture, divide the document into different layout areas, and return the type, location, content, and confidence of each area.
characteristic
- Easy to use: through the segmentation results of annotation areas with different colors, it supports the two-way rapid positioning of original annotation and visual results.
- Multiple formats: support JPEG, JPG, PNG, PDF, BMP, TIFF format and other documents.
usage method
Text version
New model
1 log in to mage platform from the following pathPre training AI ability / picture understanding / layout analysisEnter the layout analysis model page.
2. Create a new layout analysis model and select OCR engine according to requirements.
Layout analysis test
1 select the layout analysis model and click the upper right cornerLayout analysis test, enter the test page of the model.
2 upload a picture and clickStart testing, get the test results.
- Support content filtering based on region type
- Click to download the content of the current test results
- Click JSON results to view more detailed results
Area type
The layout analysis model supports the following 19 types of analysis:
| Serial number | Type name | key |
|---|---|---|
| one | Document title | article_ title |
| two | Hierarchy title | paragraph_ heading |
| three | paragraph | paragraph |
| four | Table title | table_ title |
| five | Full frame table | full_ line_ table |
| six | Missing line table | wireless_ table |
| seven | Tabular Note | table_ annotation |
| eight | Picture title | picture_ title |
| nine | picture | picture |
| ten | Photo Annotate | picture_ annotation |
| eleven | Page Notes | page_ annotation |
| twelve | header | page_ header |
| thirteen | footer | page_ footer |
| fourteen | list | list |
| fifteen | seal | seal |
| sixteen | QR code | qr_ code |
| seventeen | autograph | handwritten_ signature |
| eighteen | Logo | logo |
| nineteen | other | other |