Eyeblink

Eyeblink is an eye-blink tracking model that consumes a cropped image of eyes and predicts the eye openness.

Demo: https://mirrory-dev.github.io/eyeblink/

This is based on Taehee Lee's Eye Blink Detector.

Usage

yarn add https://github.com/mirrory-dev/eyeblink

Multi-architecture support

Node.js

Load tf.js Model from file-system and use Tensorflow C binary to predict.

Run yarn add @tensorflow/tfjs-node

import * as eyeblinkModel from '@mirrory/eyeblink';
const model = await eyeblinkModel.load('path/to/tfjs-model');

Browser

Load tf.js Model from URL and use Tensorflow.js to predict.

Run yarn add @tensorflow/tfjs

import * as eyeblinkModel from '@mirrory/eyeblink/dist/umd/eyeblink';
const model = await eyeblinkModel.load('https://path/to/tfjs-model');

Dev

./train.sh

yarn install
yarn bootstrap
yarn link

cd demo
yarn install
yarn start
open https://localhost:5000