In today's scenario LaTex is used in almost all the field of study from writing a small lab report to publishing large research paper. LaTex is similar to code so many people sometimes find it tough to write in LaTex especially when it comes to writing a complex mathematical equation. Keeping this in mind we have developed a model that would create a LaTex code of any mathematical equation given it's handwritten image.
- Preprocessed the images and their LaTeX labels, from corresponding inkml files
- Trained an Encoder-Decoder model with EfficientNet-B6 as encoder and LSTM Cell with Attention as decoder.
- Achieved a Top-5 accuracy of 95% and Top-1 accuracy of 87% and deployed the model using Flask.
Steps to run this on your local computer:
- Clone this repository
git clone https://github.com/Umang1815/Math-2-Tex
- Make a new virtual environment in python in the folder in which this repository is saved and Active the Environment.
pip install virtualenv
python -m venv <myenvname>
path\to\venv\Scripts\Activate.ps1 (Run this line with your path to activate the virtual environment)
- Download the requirements of the environment using
pip install -r requirements.txt
-
Download the model weights and place them into "model_data" folder. You can get the model weights here.
-
Now, run app.py file to open the website
flask run
- Upload an image of handwritten mathematical expression
- Click on "Convert" to get the LaTex code
- Click on "Copy LaTex" to get the corresponding LaTex code
