Table Transformer Structure Recognition v1.1
| Property | Value |
|---|---|
| Parameter Count | 28.8M |
| License | MIT |
| Paper | View Paper |
| Downloads | 577,057 |
| Tensor Type | F32 |
What is table-transformer-structure-recognition-v1.1-all?
The Table Transformer (TATR) is a specialized model designed for table structure recognition in documents. Developed by Microsoft, it builds upon the DETR (Detection Transformer) architecture and has been trained on comprehensive datasets including PubTables1M and FinTabNet.c. This version implements the "normalize before" approach, applying layer normalization before self- and cross-attention operations.
Implementation Details
Based on the Transformer architecture, this model incorporates object detection capabilities specifically optimized for table recognition. The model utilizes 28.8M parameters and operates with F32 tensor precision.
- Implements DETR-based architecture with modified normalization
- Trained on specialized table recognition datasets
- Optimized for document table structure analysis
Core Capabilities
- Table structure detection in documents
- Complex table layout recognition
- Integration-ready with inference endpoints
- Robust object detection for tabular data
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its specialized focus on table structure recognition, implementing a modified DETR architecture with "normalize before" setting, making it particularly effective for document analysis tasks.
Q: What are the recommended use cases?
The model is ideal for automated document processing systems, academic paper analysis, financial document processing, and any application requiring accurate table structure recognition in digital documents.
