The aim of this project is to create an open source C++ solution for face rigging. The solution should be minimalistic and fast, modular, cross-platform, with simple API.
The project currently includes the Input Pipeline that retrieves images from a camera, detects faces and determines a head pose of dominant face using facial landmark detection. Further, there is lightweight render based on Phong reflection model. Models are loaded via Assimp therefore various object formats are supported.
Application rigs detected head pose to head model. The render has rather general-purpose nature but a facade for face rigging is defined. See the examples below.
-
OpenCV for face detection.
-
Dlib for facial landmark detection. (Will be downloaded & compiled automatically by CMake.)
-
OpenGl for render implementation.
You are most likely interested in FaceRigNoGUI target, hence do:
cmake . && make
bin/FaceRigNoGUI
Target FaceRig includes Qt based UI but for the time being
it is just camera with landmark detection.
There will be two ways how to use project code. First as library through input/output streams. Second as stand-alone application that process webcam input and routes rendered output into virtual device that can be set as default. This will allow to stream rigged face into application like Skype.
Application should also include:
-
Mesh Transformation: Possibility to map facial landmark points to model bones and transform the mesh according to weights of vertices.
-
Gaze detection: Eyes rigging.
-
Configuration via UI: Ability to load different models, choose scene and its lighting and more during runtime via user friendly UI.