Texteller
TexTeller is an end-to-end formula recognition model based on the ViT architecture, capable of recognizing mathematical formulas in natural images and converting them into LaTeX format.
Downloads 3,806
Release Time : 2/10/2024
Model Overview
TexTeller is an image-to-text model specifically designed to recognize mathematical formulas in natural images and convert them into LaTeX format. The model is trained on a larger-scale image-formula paired dataset, demonstrating excellent generalization capabilities and high accuracy.
Model Features
Large-scale training data
Training data expanded to 7.5 million entries, 15 times the previous version, with significantly improved data quality.
Outstanding performance
Demonstrates outstanding performance in test sets, particularly in recognizing rare symbols, complex multi-line formulas, and matrices.
High generalization capability
Compared to LaTeX-OCR, which uses about 100,000 data entries, it shows superior generalization capability and higher accuracy.
Model Capabilities
Image to text
Mathematical formula recognition
LaTeX format conversion
Complex symbol recognition
Multi-line formula processing
Matrix recognition
Use Cases
Academic research
Paper formula extraction
Extract mathematical formulas from academic papers or books and convert them into LaTeX format.
Improves formula entry efficiency and reduces manual input errors.
Education
Teaching material creation
Quickly convert handwritten or printed mathematical formulas into editable LaTeX format.
Simplifies the process of creating teaching materials.
Featured Recommended AI Models
Š 2025AIbase