Web application that recognizes hand-drawn digits
Uses a custom from-scratch neural net written in Rust and compiled to WebAssembly
(all snippets start in root of repo)
- Build wasm for node
cd ./rust-neural-network
wasm-pack build --target node- Unpack data
cd ./trainer
unzip ./mnist.zip- Train model
cd ./trainer
yarn train- Copy model to web app
cp ./trainer/model.json ./frontend/public- Build wasm for browser
cd ./rust-neural-network
wasm-pack build- Start website
cd ./frontend
yarn install
yarn serve