Single page web application that runs RTSort in the browser using h5wasm to stream local ephys data (.h5) files and ONNX to run the spike detetion model.
Proof of concept that runs RT-Sort pre-processing and spike detection for raw Maxell Biosystems recordings. Pre-processing is currently in javascript and could easily be moved into ONNX improving performance. The model output concordant with RT-Sort python to 3 decimal places using the CPU and 1 decimal place using the M3 GPU (see the browser console for concordance details). Spike sorting is TBD. Tested in Chrome and Safari on a MacBook M3.
To run the Vue based web application:
npm install
npm run dev
To use the scripts to export a model to ONNX and subset files:
python3.11 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
