Surya

Surya Layout Analysis¶

Here we use RT-DETR to detect analyze, its a example, and how it will ideally run. 

In [2]:

from omnidocs.tasks.layout_analysis.extractors.surya import SuryaLayoutDetector

from omnidocs.tasks.layout_analysis.extractors.surya import SuryaLayoutDetector

c:\Users\laxma\OneDrive\Desktop\CogLab\11-07-2025\Omnidocs\new\Lib\site-packages\transformers\utils\hub.py:111: FutureWarning: Using `TRANSFORMERS_CACHE` is deprecated and will be removed in v5 of Transformers. Use `HF_HOME` instead.
  warnings.warn(

In [6]:

detector = SuryaLayoutDetector(show_log=True)
image_path = "assets/news_paper.png"
annotated_image, layout_output = detector.detect(image_path)
print(f"Detected {len(layout_output.bboxes)} elements")

detector = SuryaLayoutDetector(show_log=True) image_path = "assets/news_paper.png" annotated_image, layout_output = detector.detect(image_path) print(f"Detected {len(layout_output.bboxes)} elements")

INFO     [timestamp]2025-07-29 20:16:09[/] | [logger.name]omnidocs.tasks.layout_analysis.extractors.surya[/] |     
         [function]surya.py:55[/] | [info]Initializing SuryaLayoutDetector[/]                                      

INFO     [timestamp]2025-07-29 20:16:09[/] | [logger.name]omnidocs.tasks.layout_analysis.extractors.surya[/] |     
         [function]surya.py:55[/] | [info]Initializing SuryaLayoutDetector[/]                                      

INFO     [timestamp]2025-07-29 20:16:09[/] | [logger.name]omnidocs.tasks.layout_analysis.extractors.surya[/] |     
         [function]surya.py:62[/] | [info]Using device: cuda[/]                                                    

INFO     [timestamp]2025-07-29 20:16:09[/] | [logger.name]omnidocs.tasks.layout_analysis.extractors.surya[/] |     
         [function]surya.py:62[/] | [info]Using device: cuda[/]                                                    

INFO     [timestamp]2025-07-29 20:16:09[/] | [logger.name]omnidocs.tasks.layout_analysis.extractors.surya[/] |     
         [function]surya.py:68[/] | [info]Found surya package at:                                                  
         c:\Users\laxma\OneDrive\Desktop\CogLab\11-07-2025\Omnidocs\new\Lib\site-packages\surya\__init__.py[/]     

INFO     [timestamp]2025-07-29 20:16:09[/] | [logger.name]omnidocs.tasks.layout_analysis.extractors.surya[/] |     
         [function]surya.py:68[/] | [info]Found surya package at:                                                  
         c:\Users\laxma\OneDrive\Desktop\CogLab\11-07-2025\Omnidocs\new\Lib\site-packages\surya\__init__.py[/]     

SUCCESS  [timestamp]2025-07-29 20:16:11[/] | [logger.name]omnidocs.tasks.layout_analysis.extractors.surya[/] |     
         [function]logging.py:196[/] | [success]Models initialized successfully[/]                                 

SUCCESS  [timestamp]2025-07-29 20:16:11[/] | [logger.name]omnidocs.tasks.layout_analysis.extractors.surya[/] |     
         [function]logging.py:196[/] | [success]Models initialized successfully[/]                                 

Recognizing layout: 100%|██████████| 1/1 [00:05<00:00,  5.97s/it]

INFO     [timestamp]2025-07-29 20:16:17[/] | [logger.name]omnidocs.tasks.layout_analysis.extractors.surya[/] |     
         [function]logging.py:150[/] | [info]detect completed in 6.17s[/]                                          

INFO     [timestamp]2025-07-29 20:16:17[/] | [logger.name]omnidocs.tasks.layout_analysis.extractors.surya[/] |     
         [function]logging.py:150[/] | [info]detect completed in 6.17s[/]                                          

Detected 65 elements

In [ ]:

output_path = "output/surya_result.png"
detector.visualize((annotated_image, layout_output), output_path)
print(f"Saved visualization and json to {output_path}")

output_path = "output/surya_result.png" detector.visualize((annotated_image, layout_output), output_path) print(f"Saved visualization and json to {output_path}")

Saved visualization to output/surya_result.png

In [8]:

from IPython.display import Image, display
# Display in notebook
display(Image(output_path))

from IPython.display import Image, display # Display in notebook display(Image(output_path))