Math-Parser

A lightweight modular math OCR pipeline that processes scanned equations and reconstructs them into LaTeX.

Modules

1. Symbol Classifier (CNN-based)
2. Symbol Segmentation

Datasets

For symbol classification training, we use the HASYv2 dataset.

Setup

git clone https://github.com/yourusername/math-ocr.git
cd math-ocr
pip install -r requirements.txt

Folder Structure

• data/: Store your raw, processed, and synthetic images
• models/: Trained model checkpoints
• src/: All source code modules

To download HASYv2 dataset:

import kagglehub

# Download latest version
path = kagglehub.dataset_download("guru001/hasyv2")

print("Path to dataset files:", path)

Then move files as needed.