Getting started

Python

Dynaface needs Python 3.10 or newer. Install it from PyPI:

pip install dynaface

Then measure a face in a photo:

from dynaface import facial, measures, models

# Load the AI models, using a GPU or Apple silicon when one is available.
models.init_models(models.download_models(), models.detect_device())

# Find the face, crop it to a standard size and run every measure.
face = facial.load_face_image(
    "https://data.heatonresearch.com/dynaface/sample/img4-1024-frontal.jpg",
    measures=measures.all_measures(),
)
results = face.analyze()

# results maps each measure to its value, such as results["oce.l"] in millimeters.
print(f"Facial Asymmetry Index: {results['fai']:.2f}")

# Save the cropped face with every measure drawn on it.
face.save("face_measures.jpg")

The first run downloads Dynaface's three AI models, about 400 MB, to ~/.dynaface/models. detect_device() runs them on an NVIDIA GPU ("cuda") or Apple silicon ("mps") when one is available, and on the CPU otherwise.

Next, work through the tutorial, try the examples, or browse the Python API.

.NET

Install the packages from NuGet:

dotnet add package Dynaface.Onnx

Dynaface.Onnx runs the models with ONNX Runtime and brings in the core Dynaface package. The .NET library doesn't download the models itself. Use the same three .onnx files the Python library downloads, or get them from the model bundle. The .NET page shows how to analyze a face.