Branched version of OpenFace for ROS from @interaction-lab on 12/17/16. Use this for projected face build.

This is a small ROS node that exposes OpenFace over ROS. In particular, given a image of a face, it will output:

• Eye Gaze Vectors
• Head Pose
• 2D Landmarks
• 3D Landmarks
• Action Units
• Debug Visualization (optional)

This repository expects Interaction Lab's fork of OpenFace and OpenCV 3 to be installed.

Dependencies of OpenFace include:

• dlib
• Boost Filesystem
• tbb
• OpenCV 3

• ~image_topic - The topic the image is provided on (e.g. /usb_cam/image_raw).

• ~clnf_model_path - Provide an alternate CLNF model to OpenFace.
• ~tri_model_path - Provide an alternate tri model to OpenFace.
• ~au_model_path - Provide an alternate AU model to OpenFace.
• ~haar_model_path - Provide an alternate HAAR model to OpenFace.
• ~publish_viz - Set to true to publish a debug visualization (default: false).

• Install all dependencies (see below)
• Clone Interaction Lab's fork of OpenFace, follow installation instructions, then sudo make install at the end
• Clone & re-compile cv_bridge (in vision_opencv stack)
• Clone usb_cam ros node or other ros node of your choice for interfacing with USB camera

• OpenCV 3 (If installing from source, make sure to run cmake as follows: cmake -DBUILD_SHARED_LIBS=ON ..⁠⁠⁠⁠
• dlib (You should be able to install from apt-get, if not, clone and install from source)

• roscore
• rosrun usb_cam usb_cam_node
• rosrun openface_ros openface_ros _image_topic:="/usb_cam/image_raw"

This node requires cv_bridge and OpenCV 3. You must ensure that cv_bridge is also linked against OpenCV 3. If you get a warning during compilation, you may have to manually clone the vision_opencv repository and re-build cv_bridge.

std_msgs/Header header

geometry_msgs/Vector3 left_gaze
geometry_msgs/Vector3 right_gaze

geometry_msgs/Pose head_pose

geometry_msgs/Point[] landmarks_3d
geometry_msgs/Point[] landmarks_2d

openface_ros/ActionUnit[] action_units

string name
float64 presence
float64 intensity