Face anti-spoofing detection using TensorflowJS

Contents

This demo shows how to use the Blazeface model to detect faces in a video stream. Then the face image(s) will be cropped and parsed to the classification model. The model development repo is available here and this is the Kaggle notebook to train the model.

Setup

First, you need to install Node.js and yarn on your machine, with the corresponding operating system, to build and/or run the app.

Install dependencies and prepare the build directory:

yarn

Launching the development server (by default, it's running at http://localhost:1234) :

yarn watch

Build the app:

yarn build

This will create a dist/ folder, which contains the static files needed to serve the app.

After finish building the app, you could deploy the app using Firebase hosting, using the following command.

firebase deploy

You need to login into your firebase account first and create your own hosting, refer to the firebase website for the complete guide.

Demo

We already deploy this app using a firebase hosting, and you could access it using this link.