AIbase
EN
Home
Table Transformer Structure Recognition V1.1 Pub
T

Table Transformer Structure Recognition V1.1 Pub

Developed by microsoft
A table transformer model trained on the PubTables1M dataset for table structure recognition in documents.
Text Recognition
Transformers
Open Source License:MIT#Table Structure Recognition#Document Processing#Transformer Architecture
Downloads 1,634
Release Time : 11/18/2023

Model Overview

Table Transformer (TATR) is a Transformer-based object detection model specifically designed for recognizing table structures in documents. The model is based on the DETR architecture and adopts a 'pre-normalization' setup, applying layer normalization before self-attention and cross-attention mechanisms.

Model Features

Transformer-based Architecture
Utilizes the DETR architecture, combining the powerful capabilities of Transformers for table structure recognition.
Pre-normalization Setup
Applies layer normalization before self-attention and cross-attention mechanisms to enhance model performance.
Large-scale Pretraining
Pretrained on the PubTables1M dataset, equipped with robust table recognition capabilities.

Model Capabilities

Table Detection
Table Structure Recognition
Document Analysis

Use Cases

Document Processing
Table Data Extraction
Extract table data from scanned documents or PDFs
Accurately identifies table structure and content
Document Digitization
Convert tables in paper documents into digital formats
Improves document processing efficiency
Data Management
Database Entry
Automatically recognize document tables and input them into databases
Reduces manual data entry workload
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
English简体中文繁體中文にほんご
© 2025AIbase