A free, intelligent software designed to recognize mathematical formulas within images and convert them into editable LaTeX format. It benefits researchers, students, and anyone needing to digitize mathematical content from visual sources, offering a convenient alternative to manual transcription.

How It Works

The project leverages the UniMERNet multimodal model for image-to-formula recognition. After downloading the necessary unimernet_small model, the software processes input images to identify mathematical expressions, converting them into LaTeX code. This approach aims to automate the tedious process of recreating complex formulas from documents or screenshots.

Quick Start & Requirements

• Installation:
  • Windows: Download executable; requires installation in a non-Chinese path.
  • macOS (ARM): Download executable or install via Homebrew: brew install freetex.
• Prerequisites:
  • Source: Python environment (uv sync), git lfs for model download.
  • macOS: Potential openai module issue when using multimodal models (see issue #44).
• Model: Download unimernet_small from Hugging Face and place it in the models directory.
• Running: Execute python main.py after setup.
• Docs: Video demonstration and tutorial available (link implied but not provided).

Highlighted Details

• Free and open-source intelligent formula recognition.
• Converts image-based math formulas to editable LaTeX.
• Supports Windows and macOS (ARM) platforms.
• Developed using UniMERNet, PyQt-Fluent-Widgets, and KaTeX.

Maintenance & Community

No specific details regarding maintainers, community channels (like Discord/Slack), or roadmap are provided in the README. Development appears primarily focused on macOS environments, with a separate win branch suggested for Windows users.

Licensing & Compatibility

The specific open-source license is not stated in the provided README. This omission requires clarification regarding usage restrictions, particularly for commercial applications or derivative works.

Limitations & Caveats

The macOS version may encounter an openai module error with multimodal models. Windows users must install the software outside of Chinese-language directory paths to ensure proper function. Development is primarily conducted on macOS, potentially leading to platform-specific issues or delays in Windows support.