Python library

The dynaface package does all of Dynaface's analysis. It runs on Python 3.10 or newer, on Linux, macOS and Windows.

pip install dynaface

Getting started shows how to measure a face, and the tutorial walks through it step by step.

The main pieces

dynaface.models	Download the AI models, load them and choose a device.
load_face_image()	Load a photo from a file or URL.
AnalyzeFace	Analyze a face: find the landmarks, crop it, run the measures and draw them.
all_measures()	Every measure, ready to pass to AnalyzeFace.

Analyzing an image you already have

load_face_image() reads files and URLs. For an image already in memory, such as a video frame, create an AnalyzeFace and pass the image to load_image() as an RGB array:

import cv2
from dynaface import facial, measures, models

models.init_models(models.download_models(), models.detect_device())

face = facial.AnalyzeFace(measures=measures.all_measures())
frame = cv2.cvtColor(cv2.imread("frame.png"), cv2.COLOR_BGR2RGB)  # OpenCV loads BGR
face.load_image(frame, crop=True)
results = face.analyze()

Adding a measure

Measures are plugins. Subclass MeasureBase, declare one MeasureItem per output in the constructor, and implement abbrev() and calc(). calc() must return every output it declares, whether or not its item is enabled; an item's enabled flag only controls drawing. To use the new measure, include it in the measures list you give AnalyzeFace.