LayoutLM for Invoices
| Property | Value |
|---|---|
| Parameter Count | 128M |
| License | CC-BY-NC-SA-4.0 |
| Author | magorshunov |
| Framework | PyTorch |
What is layoutlm-invoices?
Layoutlm-invoices is a specialized document question-answering model built on the multi-modal LayoutLM architecture. Fine-tuned specifically for processing invoices and general documents, this model combines visual and textual understanding to extract information accurately from complex document layouts.
Implementation Details
The model leverages a 128M parameter architecture and has been fine-tuned on a proprietary invoice dataset, along with SQuAD2.0 and DocVQA datasets. It uses both I64 and F32 tensor types and implements safetensors for improved security and efficiency.
- Multi-modal architecture combining text and layout understanding
- Fine-tuned on multiple datasets for robust performance
- Supports non-consecutive token extraction
- Implements PyTorch framework with safetensors
Core Capabilities
- Advanced document question-answering on invoices
- Non-consecutive token extraction for complex fields
- Handles multi-line address and field extraction
- Processes both PDF and image-based documents
Frequently Asked Questions
Q: What makes this model unique?
This model's ability to extract non-consecutive tokens sets it apart from traditional QA models. This is particularly useful for processing complex documents where relevant information may be spread across different areas of the page.
Q: What are the recommended use cases?
The model is specifically designed for invoice processing, document analysis, and information extraction from structured documents. It's particularly effective for tasks requiring understanding of both textual content and spatial layout.
