MediaPipe Holistic is a ready-made perception pipeline from Google's open-source MediaPipe framework that combines three models in a single pass: body pose estimation (33 landmarks), tracking of both hands (21 landmarks each) and a face mesh (468 landmarks). This yields a unified topology of 543 landmarks for the whole human figure, computed in real time on-device.
The pipeline is multi-stage: a pose model first locates the figure and regions of interest (face, hands), then dedicated models refine the landmarks for each region. It runs on CPU and GPU and, according to Google, reaches near real-time performance of roughly 18–20 frames per second on mid-tier smartphones.
MediaPipe Holistic shipped with APIs for Python, JavaScript (web), Android and iOS under the Apache 2.0 license. It is used for sign language recognition, gesture analysis, fitness apps and augmented reality, among others. The solution belongs to the MediaPipe Legacy Solutions — Google ended their support on March 1, 2023, pointing to MediaPipe Tasks as the successor.
A Perception Stack encompasses the software layers that process data from cameras, LiDARs, IMUs, microphones, and other sensors in order to recognise the surrounding environment, perform localisation, detect and track objects, and interpret the scene. It is typically the first processing stage in an autonomous robot's data pipeline, feeding its outputs to planning and control stacks.
A software library exposing APIs to build, deploy or execute machine-learning models within applications.
Used in sign language recognition pipelines, gesture analysis, fitness apps and augmented reality (AR).
A mobile operating system based on the Linux kernel, developed by Google.
Apple's mobile operating system for iPhone and iPad devices.
A family of open-source operating systems based on the Linux kernel, common on servers, in the cloud and in robotics.
Apple's operating system for Mac computers.
Microsoft's family of operating systems for personal computers and servers.
Requires a camera image stream. Google does not publish formal minimum hardware requirements; the solution runs in near real time (18–20 FPS) on mid-tier smartphones, with optional GPU acceleration.
The Python package manager. The onnxruntime package is distributed via PyPI.
The package manager for the JavaScript/Node.js ecosystem. The onnxruntime-web and onnxruntime-node packages are distributed via npm.
A 64-bit CPU architecture compatible with the x86 instruction set, dominant in PCs and servers.
A 64-bit ARM CPU architecture, common in mobile, embedded and increasingly server devices.
Local deployment on an edge or embedded device, without cloud dependency.
Deployment on mobile devices (iOS, Android).
Execution in a web browser (e.g. via WebAssembly / WebGPU).
License family: Permissive
Classified as a MediaPipe Legacy Solution; Google ended support and pointed to MediaPipe Tasks as the successor. Code remains available on an as-is basis.
Launch of the solution combining pose, hands and face in a single on-device pipeline (543 landmarks).
