Introduction

react-native-mediapipe-posedetection brings Google’s MediaPipe Pose Landmarker to React Native. It detects people in a camera feed or an image and returns 33 landmarks for each person: points on the face, arms, hands, torso, legs and feet.

How it works

1. You create a detector with usePoseDetection, giving it a MediaPipe model file and some options.
2. The hook returns a VisionCamera frame processor. You attach it to a <Camera> (or use the bundled <MediapipeCamera>).
3. Every camera frame is passed straight to MediaPipe in native code (Swift on iOS, Kotlin on Android).
4. Results come back to JavaScript through your onResults callback, along with a helper that converts landmark coordinates into screen coordinates for drawing.

For photos you don’t need a camera at all: call PoseDetectionOnImage(path, model) and await the result.

Requirements

React Native	0.74 or newer
Architecture	New Architecture only (TurboModules)
iOS	The minimum iOS version of your React Native version
Android	API 24+
Expo	SDK with New Architecture enabled, using a development build (not Expo Go)
Peer dependencies	react-native-vision-camera v4, react-native-worklets-core

New Architecture required

This library is a TurboModule and does not work on the old bridge architecture. New Architecture is on by default in React Native 0.76+ and Expo SDK 52+.

Credits

The pose detection code is based on react-native-mediapipe by cdiddy77, rewritten as a TurboModule for the New Architecture.

Next

Head to Installation.