npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2024 – Pkg Stats / Ryan Hefner

pose-capture

v0.4.0

Published

A set of components to perform Tensorflow.js pose detection on hands, face, and body. Provides a player to perform visualization overlays while source video or webcam is playing.

Downloads

9

Readme

Releases

  • 0.4.0 Added "isimage" attribute to pose detection components to allow for image sources

Pose-Capture

Pose Capture is a package which contains a number of custom elements related to capturing and recording poses in various ways, all enabled by Tensorflow.js.

Each of these components act like a normal video player, but allow pose tracking, eventing, and overlay visualization of the points.

These pose capture components can record keyframes to a JSON file, and offer the option of recording audio from the video source to allow playback synced with the keyframe points.

Video sources can be a video file supported by your browser or your webcam. As video data needs to be analyzed by Tensorflow.js, they are subject to the same security restrictions as any video being captured to a <canvas> element. This typically means that your video can't come from a different domain and must be done over HTTPS.

Most of the components DO deliver a depth coordinate to compliment the X/Y coordinates, however as the first goal of this package is to overlay points accurately over video, there hasn't been much thought put into how to normalize/use this 3rd dimension. Additionally, some of these solutions can deliver full 3D keypoints as well, but these are in meters, and don't map well to pixels on your screen.

Hopefully the 3D story will be fleshed out in future updates.

Main Components

The main components in this library serve to show video playback from a source, whether a browser supported video file or a user's webcam.

Without the helper components listed below, these components will not show visualization or a playback controls UI.

<posedetection-video>

<posedetection-video> is to capture full-body poses. The currently configured Pose Detection solution uses BlazePose.

MediaPipe BlazePose can detect 33 keypoints, in addition to the 17 COCO keypoints, it provides additional keypoints for face, hands and feet.

<bodypix-video>

<bodypix-video> is to capture full-body poses. The currently configured BodyPix solution tracks similar points as the <posedetection-video> above, but seems a bit less accurate, at least the way it's configured in this package. BodyPix seems more appropriate for live segmentation, so the use-case presented here might not be appropriate for how it was intended. More exploration is needed, so unless you know what you're doing, you may want to rely on <posedetection-video> instead.

<facelandmark-video>

<facelandmark-video> is to capture many points to create something that can be used as a 3D mesh of a person's face.

This component uses MediaPipe Facemesh as the engine.

<handpose-video>

<handpose-video> is to capture points along a person's hand including along each of their fingers.

This component uses MediaPipe Hands as the engine.

<pose-player>

<pose-player> is to playback keyframes and audio captured by the above components

Helper Components

Helper components are designed to live as children inside the main pose capture components. They adhere to a specific API and live as slotted elements, where the parent component passes playback state or keyframes to them.

These components have a specific look/style, that can be customized a small bit, but the intention is for consumers of this package to create their own using either the PlayerState interface for creating UI controls, or the AbstractVisualizer class for visualization.

When adding both of these as children, be careful about order. A visualizer that takes up the entire size of the component can block mouse input to the player controls.

<pose-playback-controls>

<pose-playback-controls> is a component designed to be a child inside an above pose capture, and offers playback and recording controls for a video.

<visualization-canvas>

<visualization-canvas> is a component designed to be a child inside an above pose capture, and offers an overlay layer to see points captured on a given keyframe

Setup

Install dependencies:

npm i

Build

This project uses the TypeScript compiler to produce JavaScript that runs in modern browsers.

To build the JavaScript version of your component:

npm run build

To watch files and rebuild when the files are modified, run the following command in a separate shell:

npm run build:watch

Both the TypeScript compiler and lit-analyzer are configured to be very strict. You may want to change tsconfig.json to make them less strict.

Testing

Note: There are no actual tests to run in this project yet

This project uses modern-web.dev's @web/test-runner for testing. See the modern-web.dev testing documentation for more information.

Tests can be run with the test script, which will run your tests against Lit's development mode (with more verbose errors) as well as against Lit's production mode:

npm test

For local testing during development, the test:dev:watch command will run your tests in Lit's development mode (with verbose errors) on every change to your source files:

npm test:watch

Alternatively the test:prod and test:prod:watch commands will run your tests in Lit's production mode.

Dev Server

This sample uses modern-web.dev's @web/dev-server for previewing the project without additional build steps. Web Dev Server handles resolving Node-style "bare" import specifiers, which aren't supported in browsers. It also automatically transpiles JavaScript and adds polyfills to support older browsers. See modern-web.dev's Web Dev Server documentation for more information.

To run the dev server and open the project in a new browser tab:

npm run serve

There is a development HTML file located at /dev/index.html that you can view at http://localhost:8000/dev/index.html. Note that this command will serve your code using Lit's development mode (with more verbose errors). To serve your code against Lit's production mode, use npm run serve:prod.

Linting

Linting of TypeScript files is provided by ESLint and TypeScript ESLint. In addition, lit-analyzer is used to type-check and lint lit-html templates with the same engine and rules as lit-plugin.

The rules are mostly the recommended rules from each project, but some have been turned off to make LitElement usage easier. The recommended rules are pretty strict, so you may want to relax them by editing .eslintrc.json and tsconfig.json.

To lint the project run:

npm run lint

Formatting

Prettier is used for code formatting. It has been pre-configured according to the Lit's style. You can change this in .prettierrc.json.

Prettier has not been configured to run when committing files, but this can be added with Husky and and pretty-quick. See the prettier.io site for instructions.

Static Site

This project includes a simple website generated with the eleventy static site generator and the templates and pages in /docs-src. The site is generated to /docs and intended to be checked in so that GitHub pages can serve the site from /docs on the master branch.

To enable the site go to the GitHub settings and change the GitHub Pages "Source" setting to "master branch /docs folder".

To build the site, run:

npm run docs

To serve the site locally, run:

npm run docs:serve

To watch the site files, and re-build automatically, run:

npm run docs:watch

The site will usually be served at http://localhost:8000.

More information

Pose Capture Documentation Site