MediaPipe — Cross-Platform ML for Live & Streaming Media

MediaPipe

MediaPipe is Google's cross-platform, customisable ML framework for live and streaming media pipelines. It ships optimised, pre-built solutions for the most common computer vision tasks and runs efficiently on-device without a server round-trip.

Key Features

• Pre-built solutions: face detection, face mesh, hand tracking, pose estimation, holistic, object detection, image segmentation, text classification
• Targets: Android, iOS, desktop (Linux/macOS/Windows), web (WebAssembly), Edge TPU
• Python, Java, Swift, Objective-C, JavaScript, and C++ APIs
• LiteRT (TFLite) runtime for low-latency inference
• MediaPipe Tasks: new unified API for model-agnostic inference
• Model Maker: fine-tune built-in solutions on custom data with a few lines of code

Quick Start

import mediapipe as mp
import cv2

mp_hands = mp.solutions.hands
hands = mp_hands.Hands(static_image_mode=False, max_num_hands=2)

cap = cv2.VideoCapture(0)
while cap.isOpened():
    ret, frame = cap.read()
    result = hands.process(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
    if result.multi_hand_landmarks:
        print(f"{len(result.multi_hand_landmarks)} hand(s) detected")

Install via ai-supply

npx ai-supply add mediapipe-cross-platform-ml-solutions

Curated mirror of the open-source MediaPipe (Apache-2.0). Get it from the source.