TeXifyMath
- #Education
The client’s goal was to create a service that uses machine learning or neural network algorithms to convert a picture of a mathematical expression, whether it is hand-written or computer-written, into a string in LaTeX format
- Machine Learning
Impact
The mobile and web-based platform allows:
- Integrated solution for recognizing handwritten mathematical formulas and converting them into digital formulas.
- Documentation (instructions) based on the investigations, model training, dataset quality verification and classification.
Services we provided
Implement image processing to extract math expressions from pictures effectively
Use advanced image processing for mathematical expression extraction.
Optimize for scalable, efficient processing of math expressions.
Tech Stack
Python
NumPy
Pandas
Pytorch
RNN
Transformers
Challenges and Solutions
🧐 Challenges
- Develop a model capable of accurately recognising and interpreting handwritten or printed mathematical expressions.
- Design the service to handle different input formats and generate the corresponding mathematical expression in LaTeX format with high accuracy.
💡 Solutions
Our solution successfully handles input photos containing handwritten mathematical formulas and generates the corresponding LaTeX string representation. The process involved the following steps:
- Extracting image features using a CNN network.
- Encoding the extracted features into an embedding. Decoding the embedding using Seq2Seq translation techniques to obtain the LaTeX string representation.
User flow
For Players: