👶
✋
✊
✌

Play Rock Paper Scissors against your computer!

We're running neural networks running entirely in your browser to recognize your plays and keep score.

Unfortunately, your browser doesn't support accessing your webcam.
Try loading this page in a modern version of Firefox or Google Chrome.

Play Rock Paper Scissors against your computer!

Unfortunately, your browser doesn't support accessing your webcam.

Try loading this page in a desktop version of Firefox or Google Chrome.

The demo is built on our GPU-accelerated TensorFire library for fully in-browser deep learning. It's fast enough to perform real-time client-side classification of live webcam video, and we're showing it off here with a cute little game.

Discuss on Hacker News.

The rock paper scissors classifier is based on the SqueezeNet architecture. TensorFire + SqueezeNet is by no means limited to this: it can identify NSFW content, distinguish 1000 different ImageNet objects, recognize gestures, detect pets, or even distinguish hot dogs from not hot dogs.

You can learn more about TensorFire and what makes it fast on the Project Page.

You can also sign up to get notified when we publish new demos or release the api documentation.

We won't spam you or give your email to third parties.

Special thanks to Simanta Gautam, Jay Palekar, Hassan Kane, Jackie Xu, Surya Bhupatiraju, Jocelyn Reyes, Laser Nite, Connor Duffy, Lily Jordan, Alexa, Caitlin, and Billy Moses for contributing short clips of shaking their fists at computers (as we all sometimes want to).