POSE SHIELD is optimized for larger screens because it uses a camera and your body movements to control the experience. We’d encourage you to visit this site again using a laptop or desktop computer. The galaxy is too big to defend on a mobile device.
Share on Twitter TensorFlow

Pose Shield

A browser-based game built with TensorFlow.js and PoseNet
Screenshot of Pose Shield game, showing a spacecraft with a multi-part force shield moving into an asteroid belt.

How to Play Pose Shield

Your spacecraft is passing through a dangerous asteroid field. As its chief shield operator, you must use your arms to activate the defenses surrounding the ship. It’s up to you and your quick reflexes to engage the right shields at the right time to protect the ship from damage.

The Power of TensorFlow.js + PoseNet

The combination of TensorFlow.js and PoseNet means you can infer real time human pose estimation using an ordinary webcam or video — all within the browser. Making this technology accessible to all opens up powerful new opportunities — everything from augmented reality to animation to gait analysis.


How does Pose Shield work?

Pose Shield uses TensorFlow.js and a pose estimation model called PoseNet to track the location of your body in real time. As your arm movements are detected in each of the seven shield zones around the spacecraft, the corresponding shields engage, blocking any incoming asteroids from doing damage.

Does Pose Shield work on my phone?

There's nothing preventing Pose Shield from working on your phone, but the small screen and camera arrangement won't create an optimal experience. We'd recommend using a laptop or desktop computer with a webcam.

Why is Pose Shield having trouble tracking my movements?

Make sure you're standing about six feet away from the camera so Pose Shield can accurately track your arm movements. If the space behind you is especially busy with furniture, signs, or other people, Pose Shield might have trouble detecting you consistently.

Note that pose estimation works best with a powerful processor and dedicated graphics hardware. Your results may vary depending on your machine.

Am I being recorded?

No. Your camera is helping PoseNet track your arm movements to control the experience, but nothing is being captured and stored.

Where can I find the code repository for this demo?

We've made the code repository publicly available on GitHub.

Is there a practical use for this technology?

Absolutely. Pose estimation has been used in a variety of industries such as assisting individuals with limited mobility, helping to prevent accidents, improving sports performance, and more.