The Table Extraction pretrained model available on ArcGIS Living Atlas of the World is a deep learning model that is used to extract tables from high-resolution scanned documents.
Extracting tables from scanned documents is essential for industries that rely on structured data, including the GIS industry. In GIS workflows, tabular data extracted from reports, survey documents, and historical records is crucial for spatial analysis, asset management, and decision making. However, many of these documents exist only as scanned images or PDFs, making it difficult to extract tables accurately. Traditional methods, such as optical character recognition (OCR) with rule-based parsing, often fail to handle complex layouts, varying table structures, and poor-quality scans, leading to errors and inefficiencies. Manual extraction is time consuming and error prone, making it inefficient for handling large volumes of documents.
The Table Extraction deep learning model addresses these challenges by using two pretrained Table Transformer models: one for detecting tables in scanned documents and another for recognizing their structure. This approach ensures table extraction with higher efficiency and accuracy. Use this model to enhance geospatial workflows by enabling seamless extraction and analysis of tabular data from scanned documents.
Model details
This model has the following characteristics:
- Input—3-band RGB imagery with a resolution better than 1000x1000 pixels.
- Output—A feature layer with boxes bounding the table (row wise) detected in the input image.
- Compute—This workflow is compute intensive and a GPU with a compute capability of 6.0 or higher is recommended.
- Architecture—This model is based on the open-source Table Transformer (TATR) models by Microsoft, which are pretrained Detection Transformer (DETR) models. The model uses two TATR models: one for table detection and another for table structure recognition.
Training Data—The TATR models have been trained on the PubTables-1M dataset by Microsoft.
Access and download the model
Download the Table Extraction pretrained model from ArcGIS Living Atlas of the World. Alternatively, access the model directly from ArcGIS Pro, or use it in ArcGIS Online using the Professional or Professional Plus user type.
Download the model from ArcGIS Living Atlas of the World
To download the model from ArcGIS Living Atlas of the World, complete the following steps:
- Browse to ArcGIS Living Atlas of the World.
- Sign in with your ArcGIS Online credentials.
- Search for Table Extraction and open the item page from the search results.
- Click the Download button to download the model.You can use the downloaded .dlpk file directly in ArcGIS Pro, or upload it and use it in ArcGIS Enterprise. Additionally, you can fine-tune the pretrained model if necessary.
Download the model using ArcGIS Pro
To download the model using ArcGIS Pro, complete the following steps:
- Open ArcGIS Pro.
- Click the Catalog pane button and select Portal.
- Click Living Atlas and search for Table Extraction.
- Right-click the model and download the .dlpk file.
Access the toolbox for post-processing the table
You can download the Postprocess Table Toolbox from ArcGIS Online using the Professional or Professional Plus user type.
Download the toolbox from ArcGIS Online
To download the toolbox, complete the following steps:
- Click the Postprocess Table Toolbox.
- Sign in with your ArcGIS Online credentials.
- Click the Download button to download the toolbox.You can use the downloaded .atbx file directly in ArcGIS Pro, or upload it and use it in ArcGIS Enterprise. Additionally, you can fine-tune the pretrained model if necessary.
Release notes
The following are the release notes:
| Date | Description |
|---|---|
| June 2025 | First release of Table Extraction |