The models used here are taken from the following repo:
https://github.com/tensorflow/tfjs-models
These examples should work in any 'modern' browser. There are no npm dependencies as the dependencies are all sourced from CDN's in the html files using script tags. Depending on your setup you may be able to double click the html files to view them in your browser.
If that doesnt work you can serve the files on any static file server. If you have npm installed you can use http-server:
npm i http-server -g
http-server -c-1 .When running for the first time it may take a while to load the model. These files can be big and will take a while to download on a slow internet connection.
Mobilenet is used for general object identification from an image. Mobilenet can distinguish between 1000 different image classes. The full list is here http://image-net.org/challenges/LSVRC/2012/browse-synsets
Try to identify what object is in a static image
Try to identify what object is visible in the webcam.
Identify and find location of multiple objects in an image. Uses COCO SSD.
Use transfer learning to train the model to identify custom objects.
Posenet is used to estimate the poses of one or multiple people in an image.
Use the webcam to find people and estimate their poses. The example has drawing helpers to superimpose the detected poses onto the webcam feed
The Speech Command Recognizer is a JavaScript module that enables recognition of spoken commands comprised of simple isolated English words from a small vocabulary
Use browser audio API to try to recognise 18 different words as well as 'background_noise' and 'unknown'. The words are 'zero', 'one', 'two', 'three', 'four', 'five', 'six', 'seven', 'eight', 'nine', 'up', 'down', 'left', 'right', 'go', 'stop', 'yes', and 'no'.
This example uses a vocabulary that recognises only the 4 directions, up, down, left and right. It has been connected up to an implementation of Snake (credit https://hjnilsson.com/).