Do you have time to waste? Try Move Mirror, the small demonstration that Google has just put online to show the progress of the present, a model capable of detecting human forms on the images by identifying the joints.
You put yourself in front of your webcam, adopt any pose, and the AI of the Mountain View firm will identify your joints. Move Mirror will match your movements in real time to hundreds of images of people doing similar poses around the world .It feels like a magic mirror that reflects your movements with images of all kinds of human activities – from sport and dance to martial arts, to theatre and beyond, explains Google. And to spice it up, the demonstration even allows you to generate a GIF from this experience.
The game is addictive, but it mainly helps to see the advances of Google in the field of computer vision. Moreover, the objective of the firm is to draw attention to the model used. "We also wanted to make automatic learning more accessible to coders and designers by introducing the estimation of the pose in the browser, which hopefully would encourage them to experiment with this technology," says Google.
If Pose Mirror is just a small game, we can already imagine how this real-time pose detection could be used in more sophisticated applications, for example to transform basic webcams into the controllers for computers.
Google also explains that this experience is based on the Tensorflow. JS Library, which allows you to launch machine learning models on the user's device. As a result, the estimation of the poses is done on the browser, and the image of the webcam is not transferred to a server.
LINK : Move Mirror
LINK : Move Mirror