Skip to content
/ TeXNet Public

im2latex solution: Transcribe mathematical expressions from screenshots using machine learning.

Notifications You must be signed in to change notification settings

rvente/TeXNet

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

74 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

The prospect of accurately transcribing mathematical expression into a markup representation is enticing because it opens the doors for bringing new life to old mathematical texts or those for which the source code is unavailable.

We make that dream a reality with this project, where we build on Sumeet Singh's work porting the model to Python3 and bringing our own dataset of 170,000 examples.

Dynamic Content

Demonstration Video

Sumeet S. Singh's Project Site

Static Content

Complete Dataset

Research Paper

Team

Directory Structure

TEXNET
│   
│      
│
└───data -> has to be made by the user (extract data folder in there)
│   │   
│   │   
│   │
│   └───formula_images
│       │   *.png    
│   
└───src -> main directory of where model will be ran
│   │   
│   │   
│   │
│   └───tools
│   │   │   prepareData.py -> run first when images are placed in inference folder
│   │   │   predict.py -> run this script after running the model based on the script outputted from prepareData.py
│   │   │   predictions.csv -> outputted csv with filenames and LaTeX predictions
│   │   
│   │   
│   │
│   └───tb_metrics -> main logging directory where logs from model training will be stored
│   │
│   │
│   │  run.py -> main file that runs the model, has settable hyperparameters and variables that need to be specified based on the training machine
|
└───predictions
│   │   
│   │   
│   │
│   └───formula_images -> paste images to perform inference on here before running inference scripts

About

im2latex solution: Transcribe mathematical expressions from screenshots using machine learning.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published