Jason Mayes

Removing people from complex backgrounds in real time using TensorFlow.js in the web browser

A Web AI Agent running entirely client side in browser, that's capable of controlling a fictional flights webpage, to get the job done by using Google's Gemma 2 (2B) model in JavaScript via WebGPU thanks to the MediaPipe Web LLM library, combined with some extra function calling logic to enable advanced user experiences.

Client side vector search using EmbeddingGemma with Web AI (LiteRT.js, TensorFlow.js, and Transformers.js)

Using headless Chrome on server side environments for true client side browser emulation with NVIDIA T4 GPUs for Web AI model testing or graphical workloads

A simple yet fast motion revealing algorithm using the live webcam feed

A small JavaScript class to programmatically position content with variable widths and heights and minimizing whitespace gaps. Imagine Pinterest on steroids (Pinterest only allows for variable height). By Jason Mayes

Fetch your twitter posts without using the new Twitter 1.1 API. Pure JavaScript! By Jason Mayes

A simple script to set up GCE environment in Google Cloud so it is ready to run TensorFlow - an Open Source Software Library for Machine Intelligence

This little utility library allows you to ask the most common question when working with video content - does the video contain something you care about? Get videos to watch themselves for you with Web AI entirely client side!

A design pattern that elegantly handles the registration of new components such that DOM upgrades are automatically performed on document load, as well as making it super easy to handle upgrades of elements that may be added after initial page load.

Retrieve large (GBs) AI binary model files from cloud, cache locally as sharded blobs to load faster on 2nd page load, returns stored file as data URL

Spotify Web AI DJ - client side agentic smarts using Gemma 2, two billion parameter LLM, to play what a user wants via natural speech input

A simple efficient configurable particle engine written in JavaScript using HTML5 canvas 2D which works on desktop and mobile by Jason Mayes

Specify a directory, this script will search it to find all MP4 videos, convert them to JPGs at a desired FPS and resize them to a given size. Super easy!

Script to convert all MP4 videos in a zip archive to JPG frames at a desired FPS with unique names. It will then retrain the top layers of a binary image classifier using TensorFlow using these extracted images.

A design pattern that elegantly handles the registration of new components such that DOM upgrades are automatically performed on document load, as well as making it super easy to handle upgrades of elements that may be added after initial page load.

A helper component to make it super easy to use handlebars templates in a re-usable manner. Instead of defining script tags inline in your HTML to define templates you can instead write "template files" which will be bought in using AJAX.

Use your Tessel with Google Compute Engine GPU instance running Darknet YOLO via a Node.js server to perform multibox detection over websockets. We can then point a servo at the object of interest and speak out what we see via the speaker.

A small library written in Google Closure to do common things fast, useful for rapid prototyping in JavaScript.

Face recognition using Tensorflow made easier and faster with Node.js + precached Python RPC server

Repo for 'Learning JavaScript Design Patterns'

Pretrained models for TensorFlow.js

A WebGL accelerated JavaScript library for training and deploying ML models.

No description.

No description.

WebGL-accelerated ML // linear algebra // automatic differentiation for JavaScript.

Short and easy to understand samples demonstrating WebGL 2 features

Google Web Starter Kit (Beta)

A gallery that showcases on-device ML/GenAI use cases and allows people to try and use models locally.

Visual Blocks for ML is a Google visual programming framework that lets you create ML pipelines in a no-code graph editor. You – and your users – can quickly prototype workflows by connecting drag-and-drop ML components, including models, user inputs, processors, and visualizations.

Stores documents used by the TensorFlow developer community

JavaScript 3D library.