Bros Base Uncased
BROS is a pre-trained language model focused on text and layout understanding, designed to efficiently extract key information from documents.
Downloads 53.22k
Release Time : 3/2/2022
Model Overview
BROS (BERT-based Representation of Spatiality) is a pre-trained language model specifically designed to process text and layout information in documents. It can extract key information from OCR results, such as ordered item lists in receipts.
Model Features
Spatial Relation Awareness
The model understands spatial layout relationships of text in documents, improving information extraction accuracy
Document Understanding Optimization
Pre-trained and optimized specifically for document information extraction tasks
OCR Result Processing
Can directly process OCR results (text + bounding boxes) as input
Model Capabilities
Document Key Information Extraction
Layout Analysis
Receipt Information Recognition
Table Data Extraction
Use Cases
Document Processing
Receipt Information Extraction
Automatically extracts itemized products, prices, etc. from scanned receipts
Generates structured data output
Table Data Extraction
Identifies and extracts table data from unstructured documents
Preserves original table structure and content relationships
Business Automation
Invoice Processing
Automatically processes invoice documents to extract key business information
Improves financial processing efficiency
Featured Recommended AI Models