Pose Playground is a playful way for users to experiment with different AI pose models and attach arbitary images to model keypoints. The main purpose of it is to enable non-developers to be playful with Machine Learning models.
- Allow users to upload images to be attached to key points in a pose tracked using a live webcam feed.
- Estimates user distance from the camera and have the object scale appropriately based on estimated distance.
- Allow the user to set translation and rotation offsets from keypoints. Offsets are maintained properly based on estimated distance.
- Object List and Property Editor.
- An option to treat white background as transparent.
Clone the repository using:
git clone https://github.com/googleinterns/posenet_playground
Cd into posenet_playground and install dependencies:
cd posenet_playground
yarn
Start the dev demo server:
yarn watch
- This repository was initially cloned from the Posenet repository. Posenet has been used to identify the various keypoints where the assets can be attached.
- The assets for the Demo.gif were taken from Creative Commons and it was recorded using screencastify.
This is not an officially supported Google product