Mathematical Formula Recognition: GLM-OCR LaTeX Output Guide

Introduction

Mathematical formula recognition is crucial for digitizing academic documents, scientific papers, and educational materials. GLM-OCR achieves state-of-the-art performance in formula recognition, outputting clean LaTeX that can be directly used in documents.

Why GLM-OCR for Formula Recognition?

GLM-OCR excels at formula recognition because:

SOTA Performance: Achieves best results on formula recognition benchmarks
LaTeX Output: Produces ready-to-use LaTeX code
Complex Formulas: Handles multi-line equations, matrices, and special symbols
Context Awareness: Understands formulas within document context

Usage

Basic Formula Recognition

# Using Ollama
ollama run glm-ocr Formula Recognition: ./equation.png

API Call

import requests

response = requests.post(
    "https://api.z.ai/api/paas/v4/layout_parsing",
    headers={"Authorization": "Bearer your-api-key"},
    json={
        "model": "glm-ocr",
        "file": "https://example.com/formula.png"
    }
)

latex_output = response.json()["content"]
print(latex_output)

Supported Formula Types

Basic Equations

Input image of: E = mc²

Output:

E = mc^2

Fractions and Roots

Input image of complex fraction:

Output:

\frac{-b \pm \sqrt{b^2 - 4ac}}{2a}

Integrals and Derivatives

\int_{0}^{\infty} e^{-x^2} dx = \frac{\sqrt{\pi}}{2}

Matrices

\begin{pmatrix}
a_{11} & a_{12} & a_{13} \\
a_{21} & a_{22} & a_{23} \\
a_{31} & a_{32} & a_{33}
\end{pmatrix}

Summations and Products

\sum_{i=1}^{n} i = \frac{n(n+1)}{2}

\prod_{i=1}^{n} i = n!

Real-World Examples

Physics Equations

Maxwell's Equations:

\nabla \cdot \mathbf{E} = \frac{\rho}{\epsilon_0}

\nabla \times \mathbf{B} = \mu_0 \mathbf{J} + \mu_0 \epsilon_0 \frac{\partial \mathbf{E}}{\partial t}

Statistical Formulas

Normal Distribution:

f(x) = \frac{1}{\sigma\sqrt{2\pi}} e^{-\frac{1}{2}\left(\frac{x-\mu}{\sigma}\right)^2}

Machine Learning

Softmax Function:

\text{softmax}(x_i) = \frac{e^{x_i}}{\sum_{j=1}^{K} e^{x_j}}

Integration with Document Workflows

LaTeX Documents

\documentclass{article}
\usepackage{amsmath}

\begin{document}

% Paste GLM-OCR output directly
The quadratic formula is:
\[
x = \frac{-b \pm \sqrt{b^2 - 4ac}}{2a}
\]

\end{document}

Markdown with MathJax

The quadratic formula is:

$$x = \frac{-b \pm \sqrt{b^2 - 4ac}}{2a}$$

Python Processing

import sympy
from sympy.parsing.latex import parse_latex

# Parse GLM-OCR LaTeX output
latex_str = r"\frac{-b + \sqrt{b^2 - 4ac}}{2a}"
expr = parse_latex(latex_str)

# Evaluate numerically
result = expr.subs({'a': 1, 'b': -5, 'c': 6})
print(float(result))  # Output: 3.0

Best Practices

Image Quality

Resolution: 200+ DPI for best results
Contrast: Black text on white background
Cropping: Isolate the formula from surrounding text

Complex Formulas

For multi-line equations:

Capture the entire equation block
GLM-OCR will preserve alignment
Output uses appropriate LaTeX environments

Handwritten Formulas

GLM-OCR also handles handwritten formulas:

Clear handwriting yields better results
Avoid overlapping symbols
Use consistent symbol sizes

Performance Tips

Batch Processing

import asyncio
import aiohttp

async def recognize_formula(session, image_url):
    async with session.post(
        "https://api.z.ai/api/paas/v4/layout_parsing",
        headers={"Authorization": "Bearer your-api-key"},
        json={"model": "glm-ocr", "file": image_url}
    ) as response:
        return await response.json()

async def batch_recognize(image_urls):
    async with aiohttp.ClientSession() as session:
        tasks = [recognize_formula(session, url) for url in image_urls]
        return await asyncio.gather(*tasks)

Comparison with Other Tools

Feature	GLM-OCR	Mathpix	Traditional OCR
Accuracy	SOTA	High	Low
LaTeX Output	Yes	Yes	No
Handwriting	Good	Good	Poor
Cost	$0.03/M tokens	$0.01/image	Varies
Local Deploy	Yes	No	Yes

Troubleshooting

Common Issues

Issue: Incorrect symbol recognition Solution: Ensure clear image with good contrast

Issue: Missing subscripts/superscripts Solution: Use higher resolution image

Issue: Alignment issues in multi-line equations Solution: Capture the complete equation block

Conclusion

GLM-OCR provides state-of-the-art formula recognition with clean LaTeX output, making it ideal for digitizing academic documents, creating educational materials, and processing scientific papers.

Introduction

Why GLM-OCR for Formula Recognition?

GLM-OCR excels at formula recognition because:

SOTA Performance: Achieves best results on formula recognition benchmarks
LaTeX Output: Produces ready-to-use LaTeX code
Complex Formulas: Handles multi-line equations, matrices, and special symbols
Context Awareness: Understands formulas within document context

Usage

Basic Formula Recognition

# Using Ollama
ollama run glm-ocr Formula Recognition: ./equation.png

API Call

import requests

response = requests.post(
    "https://api.z.ai/api/paas/v4/layout_parsing",
    headers={"Authorization": "Bearer your-api-key"},
    json={
        "model": "glm-ocr",
        "file": "https://example.com/formula.png"
    }
)

latex_output = response.json()["content"]
print(latex_output)

Supported Formula Types

Basic Equations

Input image of: E = mc²

Output:

E = mc^2

Fractions and Roots

Input image of complex fraction:

Output:

\frac{-b \pm \sqrt{b^2 - 4ac}}{2a}

Integrals and Derivatives

\int_{0}^{\infty} e^{-x^2} dx = \frac{\sqrt{\pi}}{2}

Matrices

\begin{pmatrix}
a_{11} & a_{12} & a_{13} \\
a_{21} & a_{22} & a_{23} \\
a_{31} & a_{32} & a_{33}
\end{pmatrix}

Summations and Products

\sum_{i=1}^{n} i = \frac{n(n+1)}{2}

\prod_{i=1}^{n} i = n!

Real-World Examples

Physics Equations

Maxwell's Equations:

\nabla \cdot \mathbf{E} = \frac{\rho}{\epsilon_0}

\nabla \times \mathbf{B} = \mu_0 \mathbf{J} + \mu_0 \epsilon_0 \frac{\partial \mathbf{E}}{\partial t}

Statistical Formulas

Normal Distribution:

f(x) = \frac{1}{\sigma\sqrt{2\pi}} e^{-\frac{1}{2}\left(\frac{x-\mu}{\sigma}\right)^2}

Machine Learning

Softmax Function:

\text{softmax}(x_i) = \frac{e^{x_i}}{\sum_{j=1}^{K} e^{x_j}}

Integration with Document Workflows

LaTeX Documents

\documentclass{article}
\usepackage{amsmath}

\begin{document}

% Paste GLM-OCR output directly
The quadratic formula is:
\[
x = \frac{-b \pm \sqrt{b^2 - 4ac}}{2a}
\]

\end{document}

Markdown with MathJax

The quadratic formula is:

$$x = \frac{-b \pm \sqrt{b^2 - 4ac}}{2a}$$

Python Processing

import sympy
from sympy.parsing.latex import parse_latex

# Parse GLM-OCR LaTeX output
latex_str = r"\frac{-b + \sqrt{b^2 - 4ac}}{2a}"
expr = parse_latex(latex_str)

# Evaluate numerically
result = expr.subs({'a': 1, 'b': -5, 'c': 6})
print(float(result))  # Output: 3.0

Best Practices

Image Quality

Resolution: 200+ DPI for best results
Contrast: Black text on white background
Cropping: Isolate the formula from surrounding text

Complex Formulas

For multi-line equations:

Capture the entire equation block
GLM-OCR will preserve alignment
Output uses appropriate LaTeX environments

Handwritten Formulas

GLM-OCR also handles handwritten formulas:

Clear handwriting yields better results
Avoid overlapping symbols
Use consistent symbol sizes

Performance Tips

Batch Processing

import asyncio
import aiohttp

async def recognize_formula(session, image_url):
    async with session.post(
        "https://api.z.ai/api/paas/v4/layout_parsing",
        headers={"Authorization": "Bearer your-api-key"},
        json={"model": "glm-ocr", "file": image_url}
    ) as response:
        return await response.json()

async def batch_recognize(image_urls):
    async with aiohttp.ClientSession() as session:
        tasks = [recognize_formula(session, url) for url in image_urls]
        return await asyncio.gather(*tasks)

Comparison with Other Tools

Feature	GLM-OCR	Mathpix	Traditional OCR
Accuracy	SOTA	High	Low
LaTeX Output	Yes	Yes	No
Handwriting	Good	Good	Poor
Cost	$0.03/M tokens	$0.01/image	Varies
Local Deploy	Yes	No	Yes

Troubleshooting

Common Issues

Issue: Incorrect symbol recognition Solution: Ensure clear image with good contrast

Issue: Missing subscripts/superscripts Solution: Use higher resolution image

Issue: Alignment issues in multi-line equations Solution: Capture the complete equation block

Conclusion

GLM-OCR provides state-of-the-art formula recognition with clean LaTeX output, making it ideal for digitizing academic documents, creating educational materials, and processing scientific papers.

Author

Categories

More Posts

Complex Table Recognition: How GLM-OCR Handles Merged Cells

Structured Information Extraction: GLM-OCR JSON Output for Documents

GLM-OCR Local Deployment: Ollama vs vLLM vs SGLang

Mathematical Formula Recognition: GLM-OCR LaTeX Output Guide

Author

Categories

More Posts

Complex Table Recognition: How GLM-OCR Handles Merged Cells

Structured Information Extraction: GLM-OCR JSON Output for Documents

GLM-OCR Local Deployment: Ollama vs vLLM vs SGLang