Need help with OpenSeeFace?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

180 Stars 27 Forks BSD 2-Clause "Simplified" License 256 Commits 5 Opened issues


Robust realtime face and facial landmark tracking on CPU with Unity integration

Services available


Need anything else?

Contributors list

# 178,513
244 commits



Note: This is a tracking library, not a stand-alone avatar puppeteering program. I'm also working on such an application, which is currently in a public beta.

This project implements a facial landmark detection model based on MobileNetV3.

As Pytorch 1.3 CPU inference speed on Windows is very low, the model was converted to ONNX format. Using onnxruntime it can run at 30 - 60 fps tracking a single face. There are four models, with different speed to tracking quality trade-offs.

If anyone is curious, the name is a silly pun on the open seas and seeing faces. There's no deeper meaning.

An up to date sample video can be found here, showing the default tracking model's performance under different noise and light levels.


A sample Unity project for VRM based avatar animation can be found here.

The face tracking itself is done by the
Python 3.7 script. It is a commandline program, so you should start it manually from cmd or write a batch file to start it. If you downloaded a release and are on Windows, you can run the
inside the
folder without having Python installed. You can also use the
inside the
folder for a basic demonstration of the tracker.

The script will perform the tracking on webcam input or video file and send the tracking data over UDP. This design also allows tracking to be done on a separate PC from the one who uses the tracking information. This can be useful to enhance performance and to avoid accidentially revealing camera footage.

The provided

Unity component can receive these UDP packets and provides the received information through a public field called
. The
component can visualize the landmark points of a detected face. It also serves as an example. Please look at it to see how to properly make use of the
component. Further examples are included in the
folder. The UDP packets are received in a separate thread, so any components using the
field of the
component should first copy the field and access this copy, because otherwise the information may get overwritten during processing. This design also means that the field will keep updating, even if the
component is disabled.

Run the python script with

to learn about the possible options you can set.
python --help

A simple demonstration can be achieved by creating a new scene in Unity, adding an empty game object and both the

components to it. While the scene is playing, run the face tracker on a video file:
python --visualize 3 --pnp-points 1 --max-threads 4 -c video.mp4

This way the tracking script will output its own tracking visualization while also demonstrating the transmission of tracking data to Unity.

The included

component allows starting the face tracker program from Unity. It is designed to work with the pyinstaller created executable distributed in the binary release bundles. It provides three public API functions:
  • public string[] ListCameras()
    returns the names of available cameras. The index of the camera in the array corresponds to its ID for the
    field. Setting the
    will disable webcam capturing.
  • public bool StartTracker()
    will start the tracker. If it is already running, it will shut down the running instance and start a new one with the current settings.
  • public void StopTracker()
    will stop the tracker. The tracker is stopped automatically when the application is terminated or the
    object is destroyed.


component uses WinAPI job objects to ensure that the tracker child process is terminated if the application crashes or closes without terminating the tracker process first.

Additional custom commandline arguments should be added one by one into elements of

array. For example
-v 1
should be added as two elements, one element containing
and one containing
, not a single one containing both parts.

The included

component can be used in conjunction with FinalIK or other IK solutions to animate head motion.

Expression detection


component can be added to the same component as the
component to detect specific facial expressions. It has to be calibrated on a per-user basis. It can be controlled either through the checkboxes in the Unity Editor or through the equivalent public methods that can be found in its source code.

To calibrate this system, you have to gather example data for each expression. If the capture process is going too fast, you can use the

option to slow it down.

The general process is as follows:

  • Type in a name for the expression you want to calibrate.
  • Make the expression and hold it, then tick the recording box.
  • Keep holding the expression and move your head around and turn it in various directions.
  • After a short while, start talking while doing so if the expression should be compatible with talking.
  • After doing this for a while, untick the recording box and work on capturing another expression.
  • Tick the train box and see if the expressions you gathered data for are detected accurately.
  • You should also get some statistics in the lower part of the component.
  • If there are issues with any expression being detected, keep adding data to it.

To delete the captured data for an expression, type in its name and tick the "Clear" box.

To save both the trained model and the captured training data, type in a filename including its full path in the "Filename" field and tick the "Save" box. To load it, enter the filename and tick the "Load" box.


  • A reasonable number of expressions is six, including the neutral one.
  • Before starting to capture expressions, make some faces and wiggle your eyebrows around, to warm up the feature detection part of the tracker.
  • Once you have a detection model that works decently, when using it take a moment to check all the expressions work as intended and add a little data if not.

General notes

  • The tracking seems to be quite robust even with partial occlusion of the face, glasses or bad lighting conditions.
  • The highest quality model is selected with
    --model 3
    , the fastest model with the lowest tracking quality is
    --model 0
  • Lower tracking quality mainly means more rigid tracking, making it harder to detect blinking and eyebrow motion.
  • Depending on the frame rate, face tracking can easily use up a whole CPU core. At 30fps for a single face, it should still use less than 100% of one core on a decent CPU. If tracking uses too much CPU, try lowering the frame rate. A frame rate of 20 is probably fine and anything above 30 should rarely be necessary.
  • When setting the number of faces to track to a higher number than the number of faces actually in view, the face detection model will run every
    frames. This can slow things down, so try to set
    no higher than the actual number of faces you are tracking.


Four pretrained face landmark models are included. Using the

switch, it is possible to select them for tracking. The given fps values are for running the model on a single face video on a single CPU core. Lowering the frame rate would reduce CPU usage by a corresponding degree.
  • Model -1: This model is for running on toasters, so it's a very very fast and very low accuracy model. (213fps without gaze tracking)
  • Model 0: This is a very fast, low accuracy model. (68fps)
  • Model 1: This is a slightly slower model with better accuracy. (59fps)
  • Model 2: This is a slower model with good accuracy. (50fps)
  • Model 3 (default): This is the slowest and highest accuracy model. (44fps)

FPS measurements are from running on one core of my CPU.

Pytorch weights for use with
can be found here.





More samples: Results3.png, Results4.png

Face detection

The landmark model is quite robust with respect to the size and orientation of the faces, so the custom face detection model gets away with rougher bounding boxes than other approaches. It has a favorable speed to accuracy ratio for the purposes of this project.


Release builds

The builds in the release section of this repository contain a

inside a
folder that was built using
and contains all required dependencies.

To run it, at least the

folder has to be placed in the same folder as
. Placing it in a common parent folder should work too.

When distributing it, you should also distribute the

folder along with it to make sure you conform to requirements set forth by some of the third party libraries. Unused models can be removed from redistributed packages without issue.

The release builds contain a custom build of ONNX Runtime without telemetry.


  • Python 3.7
  • ONNX Runtime
  • OpenCV
  • Pillow
  • Numpy

The required libraries can be installed using pip:

 pip install onnxruntime opencv-python pillow numpy


Training dataset

The model was trained on a 66 point version of the LS3D-W dataset.

  title={How far are we from solving the 2D \& 3D Face Alignment problem? (and a dataset of 230,000 3D facial landmarks)},
  author={Bulat, Adrian and Tzimiropoulos, Georgios},
  booktitle={International Conference on Computer Vision},

Additional training has been done on the WFLW dataset after reducing it to 66 points and replacing the contour points and tip of the nose with points predicted by the model trained up to this point. This additional training is done to improve fitting to eyes and eyebrows.

  author = {Wu, Wayne and Qian, Chen and Yang, Shuo and Wang, Quan and Cai, Yici and Zhou, Qiang},
  title = {Look at Boundary: A Boundary-Aware Face Alignment Algorithm},
  booktitle = {CVPR},
  month = June,
  year = {2018}

For the training the gaze and blink detection model, the MPIIGaze dataset was used. Additionally, around 125000 synthetic eyes generated with UnityEyes were used during training.

The heatmap regression based face detection model was trained on random 224x224 crops from the WIDER FACE dataset.

  Author = {Yang, Shuo and Luo, Ping and Loy, Chen Change and Tang, Xiaoou},
  Booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  Title = {WIDER FACE: A Face Detection Benchmark},
  Year = {2016}


The algorithm is inspired by:

The MobileNetV3 code was taken from here.

For all training a modified version of Adaptive Wing Loss was used.

For expression detection, LIBSVM is used.

Face detection is done using a custom heatmap regression based face detection model or RetinaFace.

  title={RetinaFace: Single-stage Dense Face Localisation in the Wild},
  author={Deng, Jiankang and Guo, Jia and Yuxiang, Zhou and Jinke Yu and Irene Kotsia and Zafeiriou, Stefanos},

RetinaFace detection is based on this implementation. The pretrained model was modified to remove unnecessary landmark detection and converted to ONNX format for a resolution of 640x640.


Many thanks to everyone who helped me test things!


The code and models are distributed under the BSD 2-clause license.

You can find licenses of third party libraries used for binary builds in the


We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.