applied ml
Handwritten Equation Solver
A custom CNN that reads handwritten math and solves it: 14 symbols at 95% test accuracy.
about
An OpenCV pipeline normalizes input to 32×32 grayscale and segments individual symbols, then a custom CNN classifies 14 mathematical symbols at 95% test accuracy before the expression is parsed and evaluated. Next.js frontend on Vercel, FastAPI inference API on Hugging Face Spaces.
95%
test accuracy
14
symbol classes
stack
- TensorFlow
- OpenCV
- FastAPI
- Next.js
- Hugging Face