MediaPipe Face Mesh Plotting: Instant 3D Facial Landmarks on Any Image

What It Is
MediaPipe Face Mesh Plotting is a compact model on AIOZ AI V1 that can detect up to 468 facial landmarks from scanned images and overlay them with a precise 3D mesh.
Built on the MediaPipe Face Mesh framework, it can be integrated into Python projects or web applications easily and is ideal for face-based analysis or effects.

How It Works
- Input: User uploads a .jpg or .png image
- Processing: The model detects one or two faces and plots the full 3D landmark mesh
- Output: The model returns an annotated image and the raw landmark coordinates (.png + .json) for easy downstream use
Key Features
- Dense Landmarks: 3D points for detailed expression and pose analysis.
- Two-Face Support: Tracks up to two faces per frame by default.
- PyTorch + MediaPipe core: Runs on CPU or GPU with minimal dependencies.
Ideal Use Cases
- AR Filters & Virtual Try-On: Anchor glasses, makeup, or masks precisely.
- Avatar Animation: Drive rigged characters with real facial movement.
- Health & Sports Analytics: Measure symmetry, track fatigue, or muscle shifts.
- Security & Fintech: Add liveness detection to face ID pipelines.
Getting started
- Install the three required Python packages.
- Call the model with a photo that includes a face.
- Receive a mesh-annotated image alongside its landmark coordinates.
- Overlay, store, or analyze the results as you wish—no extra post-processing is needed.
License
Based on the original work by Hysts. Full license details are available on the model page.
Try it now
Upload a face photo and watch the mesh appear in seconds on the MediaPipe Face Mesh Plotting model page.