/laugh-api.js

[University, 3rd Year] JavaScript laughter recognition API for the browser implemented on top of tensorflow.js core

Primary LanguageTypeScriptMIT LicenseMIT

LOLd.js - Multimodal Lightweight Online Laughter Detection

JavaScript (TS) API for laughter recognition with audiovisual data, built using Tensorflow.js

The full report is available as part of this repo.

The multimodal recognition uses two different models:

  • For video data the model used is face-api.js
  • For audio data, a custom model has been built (using RNN with LSTMs)

Setup dev environment

  1. we need to transpile the .js code to .ts:
    tsc --watch
    Using the watch flag so that the files are transpiled automatically upon change.

  2. we need to serve the files:
    npm start
    This will have webpack-dev-server serve the files on localhost:8080

* * * * * * * * * * * * * * * * * * * * * * * * *
* Final Year Project (2019/2020)                *
* BSc(Eng) "Creative Computing"                 *
* Queen Mary University of London               *
*                                               *
* @Authors Alberto Morabito                     *
* * * * * * * * * * * * * * * * * * * * * * * * *