ofxFaceTracker is openFrameworks addon for face tracking, based on Jason Saragih's FaceTracker library.
This addon breaks on OF 0.11.0 due OpenCV 4. Consider using ofxFaceTracker2
openFrameworks is an open source toolkit for creative coding.
All ofxFaceTracker code is available under the MIT license, while FaceTracker is provided free for non-commercial use. For commercial use of FaceTracker, please request a quote.
- ofxTiming (for some examples)
- ofxDelauny (for some examples)
First, download ofxCv.
Then, you need to make a copy of the
libs/FaceTracker/model/ directory in
example-*/bin/data/model/ of each example. You can do this by hand, or
python setup.py will take care of this for you.
Then you can generate project files with the OF project generator and run the examples. If you see the error
Assertion failed: s.is_open() when running your app, that means you forgot to drop the model files in the right directory.
If you would like to prototype an idea involving face tracking, I encourage you to download FaceOSC. Dan Wilcox has some great FaceOSC templates that will help you get started in environments like Max, pd, Processing, and openFrameworks.
If you're interested in using ofxFaceTracker for face substitution, check out the FaceSubstitution repository.
Getting the combination of ofxFaceTracker and ofxCv working on iOS can be complicated and confusing. Try using this ofxFaceTracker-iOS repo as a starting point. Scott Hawley has also made a FaceOSC iOS port to interface with Wekinator that may helpful.
Demonstrates how to get the image-space position of the face, and the 3d orientation (collectively, the face "pose") then applies these to the OpenGL context in order to draw an oriented face mesh.
An advanced example that shows how you might detect blinking from the extracted eye images. This is not necessarily a robust classifier for whether eyes are open or closed, but works well for blinks as events.
Fairly complicated example that demonstrates some advanced/experimental features of ofxCv and ofxFaceTracker for doing AR-style augmentation of the face.
Provides a minimal example of using an
ofxVideoGrabber with an
ofxFaceTracker, then extracts and draws the
ofPolyline representing the left and right eyes.
ExpressionClassifier, which can load, save, and classify expressions into pre-trained categories. Provides basic expressions (eyebrows raised, neutral, smiling) as examples.
Demonstrates how to use the mean face mesh to draw pose and expression normalized representation of the face currently being tracked.
FaceOSC sends as much data as possible from
ofxFaceTracker via OSC, including its pose/position and gestural data, and its raw points. It also streams the image over Syphon.
You can download a pre-built binary in the releases page.
- center position: /pose/position
- scale: /pose/scale
- orientation (which direction you're facing): /pose/orientation
- mouth width: /gesture/mouth/width
- mouth height: /gesture/mouth/height
- left eyebrow height: /gesture/eyebrow/left
- right eyebrow height: /gesture/eyebrow/right
- left eye openness: /gesture/eye/left
- right eye openness: /gesture/eye/right
- jaw openness: /gesture/jaw
- nostril flate: /gesture/nostrils
- raw points (66 xy-pairs): /raw