Human Pose Detection - Machine Model Training

(This repo consists the source code of a machine model training tool.)

Demo Project (The Machine model has been trained using this repo) · Report Bug · Request Feature

About The Project

This is a sample Human Pose Detection project developed using ml5.js PoseNet. Custom machine model can be trained for the recognition of custom human poses. (Hands up, squat etc.)

In order to train a machine model, there should be raw data collected. When there is raw data ready, we can proceed to train a machine model from that data. Once the training is complete the machine model can be downloaded and consumed in any application.

Machine model training should be done with the buttons below the video element. Pose Label saves the name of the pose. Make sure you enter a valid name before starting the training for your desired human pose. Collect Data button starts the data collection for a specific human pose and snap shots the body composition in the video constantly for 20 seconds and collects raw image data. Data collection has 5 seconds delay to give you enough time to be ready in front of the camera. Status header will have informative directives.

Multiple human poses can be trained using Collect Data button. Make sure you enter Pose Label for each pose you want to train the machine model for. Once data collection is done for each human pose you want, click Train button to train a machine model from the raw image data. It will open up a side bar to show the machine model training progress. Once it is finished, click Save Machine Model button to download the machine model you trained with your desired human poses.

Now you have a custom machine model which will recognize the human poses you train it for. Feel free to build fun things with it and let me know what you build.

Prerequisites

The tools/software below are needed to build/extend this project.

• node

• npm

npm install npm@latest -g

Bundled with

• Babel
• Webpack

Editing/Extending/Usage

1. Clone the repo

git clone https://github.com/ahmettekeli/HumanPoseDetection.git

2. Install NPM packages

npm install

3. Make changes and bundle the project with webpack/babel

npm run build

Make sure you have devdependencies in package.json installed and have a "build" config under scripts section in package.json with the following config

"build": "webpack --config webpack.config.js"

webpack.config.js file with the following configs

const path = require("path");

module.exports = {
  mode: "development",
  entry: "./src/app.js",
  output: {
    path: path.resolve(__dirname, "dist"),
    filename: "app.bundle.js",
  },
  module: {
    rules: [
      {
        exclude: "/node_modules/",
        loader: "babel-loader",
      },
    ],
  },
};

Contributing

Contributions are what make the open source community such an amazing place to be learn, inspire, and create. Any contributions you make are greatly appreciated.

1. Fork the Project
2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
3. Commit your Changes (git commit -m 'Add some AmazingFeature')
4. Push to the Branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

License

Distributed under the MIT License. See LICENSE for more information.

Contact

Ahmet Tekeli - @ahmettekeli3 - ahmettekeli1991@hotmail.com

Project Link: https://github.com/ahmettekeli/HumanPoseDetection