This repository contains a demo application for pose detection using TensorFlow.js and MediaPipe Pose API. It demonstrates how to set up a basic web application to upload videos and process them for pose detection in real-time.
0.0.1
The demo provides an interface to upload video files and utilize a pose detection API to analyze and display the detected poses. The front end is built with simple HTML/CSS and JavaScript, leveraging various TensorFlow.js models and the MediaPipe Pose library.
This project requires Node.js version 8.9.0 or higher.
Clone the repository and install dependencies:
git clone <repository-url>
cd posedetection_demo_upload_video
npm installIn the project directory, you can run:
npm startRuns the app in the development mode. Open http://localhost:3000 to view it in the browser. The page will reload if you make edits.
npm run buildBuilds the app for production to the dist folder. It correctly bundles in production mode and optimizes the build for the best performance.
npm run lintLints and checks the code for any issues as per configured ESLint rules.
Build Dependencies: Compiles the necessary dependencies. Link Core: Symlinks the local TensorFlow Core package for development. Link WebGL: Symlinks the local TensorFlow WebGL backend package for development.
This project uses the following main dependencies:
@mediapipe/pose@tensorflow-models/pose-detection- TensorFlow.js modules
fs-extrafor enhanced file system operationsscatter-glfor graphical display of pose data
Includes necessary Babel and Parcel configurations for building and running the application:
@babel/core@babel/preset-envparcel-bundlereslintwith Google's ESLint configuration for code linting Other utilities like yalc for managing local package dependencies.
This demo does not support the Crypto API in the browser environment.