pose captcha

Inspiration

To create a more engaging and privacy-focused human verification system, moving beyond traditional CAPTCHAs by leveraging real-time pose detection.

What it does

PoseCaptcha is a web-based human verification system that uses a user's webcam to detect specific body poses (T-pose and Arms-up) in real-time, confirming they are human. All processing occurs locally in the browser.

How we built it

We built PoseCaptcha using Vite, React 18, and TypeScript for the frontend, styled with Tailwind CSS. Real-time pose detection is powered by TensorFlow.js with the MoveNet SINGLEPOSE_LIGHTNING model, running efficiently on the WebGL backend.

Challenges we ran into

Key challenges included optimizing TensorFlow.js performance for smooth real-time detection in the browser, ensuring accurate pose recognition across various user positions, and robustly managing the complex state and timing for the two-pose challenge.

Accomplishments that we're proud of

We successfully developed a production-ready, privacy-preserving human verification system with seamless real-time pose detection and clear visual feedback. It offers an innovative and user-friendly alternative to conventional CAPTCHAs.

What we learned

We gained significant experience in integrating machine learning models (TensorFlow.js) into web applications, optimizing their performance, and managing complex real-time state within a React environment.

What's next for Pose Captcha

Future plans include exploring more dynamic and complex pose challenges, integrating with backend systems for advanced security, and potentially adding support for different pose detection models or accessibility features.

Built With

• bolt
• netlify