MathOCR

Converts PDFs and images into tex

Dependencies

• CUDA
• CUDNN
• Boost
• Libtorch
• Torch-TensorRT
• OpenCV with Contrib(for CUDA)
• Ghostscript
• Katex
• NodeJS
• wkhtmltox
• Tesseract

Models

• YoloV8 - deskew
• OCR

Datasets

• Document Object Classification Dataset - for deskew
• Math Image to Latex Dataset

Usage

MathOCR train [dataDirectory] [epochs] [learningRate]
MathOCR normalize [inputFile] [outputDir] [deskew]
MathOCR [inputFile] [outputDir]