Dataset Viewer
Auto-converted to Parquet Duplicate
image
imagewidth (px)
2.64k
3.71k
file_name
stringclasses
3 values
markdown
stringclasses
3 values
surya_blocks
stringclasses
3 values
inference_info
stringclasses
1 value
PublicInvestmentProgram3YearRolling,FY2013-FY2015_page_010.png
<table> <tbody><tr> <th>αžŸαžΌαž…αž“αžΆαž€αžšαžŸαŸαžŠαŸ’αž‹αž€αž·αž…αŸ’αž…</th> <th>្០៑្ αž‚αžΊ</th> <th>αž†αŸ’αž“αžΆαŸ† ្០៑៣ αž‚αžΊ</th> <th>αž†αŸ’αž“αžΆαŸ†αŸ’αŸ αŸ‘αŸ€ αž‚αžΊ</th> <th>αž†αŸ’αž“αžΆαŸ† ្០៑αŸ₯ αž‚αžΊ</th></tr> <tr> <td>αž•αž›αž·αžαž•αž›αž€αŸ’αž“αž»αž„αžŸαŸ’αžšαž»αž€αžŸαžšαž»αž” (GDP) αž‚αž·αžαžαŸ’αžšαžΉαž˜αžαž˜αŸ’αž›αŸƒαž”αž…αŸ’αž…αž»αž”αŸ’αž”αž“αŸ’αž“ (αž–αžΆαž“αŸ‹αž›αžΆαž“αžšαŸ€αž›)</td> <td>αŸ₯៨,៑៩្</td> <td>៦ៀ,០៨្</td> <td>៧០,៦៧៧</td> <td>៧៧,ៀ៧៩</td></tr> <tr> <td>αž•αž›αž·αžαž•αž›αž€αŸ’αž“αž»αž„αžŸαŸ’αžšαž»αž€αžŸαžšαž»αž” (GDP)...
[{"rows": [], "cols": [], "cells": [], "image_bbox": [0.0, 0.0, 3712.0, 1680.0], "raw": "<table>\n<tbody><tr>\n<th>αžŸαžΌαž…αž“αžΆαž€αžšαžŸαŸαžŠαŸ’αž‹αž€αž·αž…αŸ’αž…</th>\n<th>្០៑្ αž‚αžΊ</th>\n<th>αž†αŸ’αž“αžΆαŸ† ្០៑៣ αž‚αžΊ</th>\n<th>αž†αŸ’αž“αžΆαŸ†αŸ’αŸ αŸ‘αŸ€ αž‚αžΊ</th>\n<th>αž†αŸ’αž“αžΆαŸ† ្០៑αŸ₯ αž‚αžΊ</th></tr>\n<tr>\n<td>αž•αž›αž·αžαž•αž›αž€αŸ’αž“αž»αž„αžŸαŸ’αžšαž»αž€αžŸαžšαž»αž” (GDP) αž‚αž·αžαžαŸ’αžšαžΉαž˜αžαž˜αŸ’αž›αŸƒαž”αž…αŸ’αž…αž»αž”αŸ’αž”αž“αŸ’αž“ (αž–αžΆαž“αŸ‹αž›αžΆαž“αžšαŸ€αž›)</td>\n<td>αŸ₯៨...
[{"model": "datalab-to/surya-ocr-2", "model_name": "surya-ocr-2", "task": "table", "table_mode": "full", "backend": "vllm-offline", "long_table_enabled": true, "long_table_threshold": 2200, "tile_height": 1500, "tile_overlap": 300, "table_max_tokens": 8192, "output_column": "markdown", "blocks_column": "surya_blocks", ...
PublicInvestmentProgram3YearRolling,FY2013-FY2015_page_011.png
<table> <tbody> <tr> <th></th> <th>αž†αŸ’αž“αžΆαŸ† ្០៑្</th> <th>αž†αŸ’αž“αžΆαŸ† ្០៑៣</th> <th>αž†αŸ’αž“αžΆαŸ† ្០៑ៀ</th> <th>αž†αŸ’αž“αžΆαŸ† ្០៑αŸ₯</th></tr> <tr> <td>αž€αžŸαž·αž€αž˜αŸ’αž˜ αž“αŸαžŸαžΆαž‘ αž“αž·αž„αž–αŸ’αžšαŸƒαžˆαžΎ</td> <td>៑,៨%</td> <td>ៀ.្%</td> <td>៣.្%</td> <td>៣.៑%</td></tr> <tr> <td>αžŠαŸ†αžŽαžΆαŸ†</td> <td>៑,៨%</td> <td>αŸ₯.្%</td> <td>៣.៦%</td> <td>៣.៣%</td></tr> <tr> <td>αžŸαžαŸ’αžœαž…αž·αž‰αŸ’αž…αžΉαž˜ αž“αž·αž„...
[{"rows": [], "cols": [], "cells": [], "image_bbox": [0.0, 0.0, 2640.0, 1500.0], "raw": "<table>\n<tbody><tr>\n<th></th>\n<th>αž†αŸ’αž“αžΆαŸ† ្០៑្</th>\n<th>αž†αŸ’αž“αžΆαŸ† ្០៑៣</th>\n<th>αž†αŸ’αž“αžΆαŸ† ្០៑ៀ</th>\n<th>αž†αŸ’αž“αžΆαŸ† ្០៑αŸ₯</th></tr>\n<tr>\n<td>αž€αžŸαž·αž€αž˜αŸ’αž˜ αž“αŸαžŸαžΆαž‘ αž“αž·αž„αž–αŸ’αžšαŸƒαžˆαžΎ</td>\n<td>៑,៨%</td>\n<td>ៀ.្%</td>\n<td>៣.្%</td>\n<td>៣.៑%</td></tr>\n<tr>...
[{"model": "datalab-to/surya-ocr-2", "model_name": "surya-ocr-2", "task": "table", "table_mode": "full", "backend": "vllm-offline", "long_table_enabled": true, "long_table_threshold": 2200, "tile_height": 1500, "tile_overlap": 300, "table_max_tokens": 8192, "output_column": "markdown", "blocks_column": "surya_blocks", ...
PublicInvestmentProgram3YearRolling,FY2013-FY2015_page_012.png
<table> <tbody> <tr> <th>αž”αŸ’αžšαž—αž–αž αž·αžšαž‰αŸ’αž‰αž”αŸ’αž”αž‘αžΆαž“</th> <th>αž†αŸ’αž“αžΆαŸ† ្០៑្</th> <th>αž†αŸ’αž“αžΆαŸ† ្០៑៣</th> <th>αž†αŸ’αž“αžΆαŸ† ្០៑ៀ</th> <th>αž†αŸ’αž“αžΆαŸ† ្០៑αŸ₯</th> <th>αžŸαžšαž»αž” αž†αŸ’αž“αžΆαŸ†αŸ’αŸ αŸ‘αŸ£ -្០៑αŸ₯</th></tr> <tr> <td>αž€αžΆαžšαžœαž·αž“αž·αž™αŸ„αž‚αžŸαžšαž»αž”</td> <td>៣,ៀ៣ៀ.៧</td> <td>៣,៨៩αŸ₯.ៀ</td> <td>ៀ,៑៧្.០</td> <td>ៀ,αŸ₯៩៧.៧</td> <td>៑្,៦៦៦.៑</td></tr> <tr> <td>Β· αž€αžΆαžšαžœαž·αž“αž·αž™αŸ„αž‚αžŸαžΆαž’αžΆαžšαžŽαŸˆ</td> <td...
[{"rows": [], "cols": [], "cells": [], "image_bbox": [0.0, 0.0, 2768.0, 1500.0], "raw": "<table>\n<tbody><tr>\n<th>αž”αŸ’αžšαž—αž–αž αž·αžšαž‰αŸ’αž‰αž”αŸ’αž”αž‘αžΆαž“</th>\n<th>αž†αŸ’αž“αžΆαŸ† ្០៑្</th>\n<th>αž†αŸ’αž“αžΆαŸ† ្០៑៣</th>\n<th>αž†αŸ’αž“αžΆαŸ† ្០៑ៀ</th>\n<th>αž†αŸ’αž“αžΆαŸ† ្០៑αŸ₯</th>\n<th>αžŸαžšαž»αž” αž†αŸ’αž“αžΆαŸ†αŸ’αŸ αŸ‘αŸ£ -្០៑αŸ₯</th></tr>\n<tr>\n<td>αž€αžΆαžšαžœαž·αž“αž·αž™αŸ„αž‚αžŸαžšαž»αž”</td>\n<td>៣,ៀ៣ៀ.៧</td>\n<td>៣,៨៩αŸ₯.ៀ<...
[{"model": "datalab-to/surya-ocr-2", "model_name": "surya-ocr-2", "task": "table", "table_mode": "full", "backend": "vllm-offline", "long_table_enabled": true, "long_table_threshold": 2200, "tile_height": 1500, "tile_overlap": 300, "table_max_tokens": 8192, "output_column": "markdown", "blocks_column": "surya_blocks", ...

Surya OCR 2 Table Recognition on sopheakvoatei/rendered_khmer_tables

Surya OCR 2 table recognition using offline vLLM inference.

The table pipeline automatically detects tall table images and applies vertical overlapping tiling before reconstructing the final HTML table.

Processing Details

Long Table Handling

Tables taller than 2200 pixels are automatically tiled.

  • Tile height: 1500 pixels
  • Tile overlap: 300 pixels
  • Table recognition: TableRecPredictor(..., mode="full")
  • Duplicate boundary rows are removed during reconstruction.
  • Repeated header rows are removed.
  • The final output is reconstructed as a single HTML table.

Output

markdown

Final table HTML.

surya_blocks

Structured Surya output containing:

  • original Surya table result
  • reconstructed HTML
  • tile count
  • tile metadata
  • individual tile results for long tables

Generated using offline vLLM inference on Hugging Face Jobs.

Downloads last month
14