Ucas-HaoranWei/GOT-OCR2.0
GOT OCR2.0
GOT-OCR2.0 is an end-to-end optical character recognition system and document text extractor. It utilizes a unified transformer architecture to recognize and extract plain and formatted text from diverse images and documents.
The system features a multi-crop processing method that divides high-resolution or dense documents into smaller sections to maintain recognition detail. It also includes a renderer that transforms recognized text into HTML to preserve the original structure and layout of the document.
The project provides a framework for fine-tuning pre-trained models on custom datasets for specialized domains. It further includes utilities for model performance evaluation and benchmarking using multi-GPU acceleration.
Features
- End-to-End Architectures - Utilizes a unified transformer architecture to process images directly into structured text sequences without multi-stage pipelines.
- High-Resolution Document OCR - Capturing high-detail text from large or dense documents by processing the image in smaller cropped sections.
- Image Tiling - Divides high-resolution images into smaller overlapping tiles to maintain pixel density for fine-grained text recognition.
- Document Text Extractors - Captures plain and formatted text from diverse document types and complex image layouts.
- Multi-Crop Processing - Divides large or dense documents into smaller sections to maintain high recognition detail.
- Text Extraction and OCR - Extracts plain and formatted text from diverse images and documents using a unified deep learning model.
- Optical Character Recognitions - Recognizes plain and formatted text from diverse document types using a unified deep learning model.
- Layout Recovery - Converts scanned images of complex documents into HTML outputs that preserve the original layout and formatting.
- Multi-Crop Processing - Captures high-detail text across large or dense documents by dividing complex images into smaller sections.
- Visual-Textual Alignments - Maps image features and text tokens into a shared latent space to correlate visual structure with linguistic meaning.
- GPU-Accelerated Inference - Implements multi-GPU acceleration to increase throughput during large-scale document processing.
- Vision Model Fine-Tuning - Provides a framework for adjusting pre-trained vision transformers using specialized datasets to improve domain-specific vocabulary recognition.
- OCR Model Fine-Tuners - Provides procedures for retraining OCR models on custom datasets to improve recognition accuracy for specialized domains.
- HTML Document Renderers - Transforms recognized text and layout coordinates into structured HTML to preserve the original document formatting.
- Data Processing - Optical character recognition model for document understanding.
- Data Processing Tools - OCR model for document understanding and text extraction.
