IRLFirst physical meetup — Bengaluru, Sat May 23, 4PM · RSVP on Luma
HomeToolsMCPHow It WorksStoriesPhilosophyCommunityArchitectureStar on GitHub
All Tools
M
Dev ToolsFreeOpen Source

MEDIAPIPE

On-device machine learning for everyone

Apache-2.0

ABOUT

Building real-time ML features like hand tracking, pose estimation, or object detection from scratch requires deep expertise in computer vision, model optimization for the target device, and platform-specific deployment code. Most teams have neither the time nor the specialized knowledge to train and ship custom vision models on mobile or web. MediaPipe solves this by providing production-ready, pre-trained models and cross-platform APIs that run entirely on-device — no cloud calls, no latency, no privacy concerns. Developers integrate a few lines of code and get real-time ML capabilities that work across Android, iOS, web, and Python with consistent APIs.

INSTALL
pip install mediapipe npm install @mediapipe/tasks-vision

INTEGRATION GUIDE

1. Add real-time hand tracking to a web AR/VR application using the web API 2. Deploy on-device object detection on Android without network connectivity or cloud costs 3. Build a pose estimation app for fitness coaching that runs at 30 FPS on mobile devices 4. Integrate face landmark detection into a video conferencing app for virtual backgrounds 5. Classify text or audio on-device for privacy-sensitive applications like health monitoring

TAGS

computer-visionon-devicemobilewebmachine-learninggooglepose-estimationobject-detection