Final project for Computer Vision course at Federal University of Alagoas.
A ROS package that recognizes faces from video input. Through proper namespaces, allows multiples nodes to run indepently.
In order to capture the video input, video_stream_opencv ROS Package was used. It allows you to used video from an external hardware (webcam) or from a stored video file.
For the recognition task, it uses the Face Recognition Python module, which is based on Dlib library.
All the recognition code is present in face_detector.py. ROS related code is present in recognition_node.py, which makes use of the former.
You can start recognition node with following command:
roslaunch ros_face_recognition start.launch \
node_name:=recon_node \
process_each_n:=3 \
camera_topic:=/camera/image_raw \
scale_factor:=0.25 \
hertz:=30
These are the default parameters. In the case you do not want to change one or more of them, just call
roslaunch ros_face_recognition start.launch
and it should start the same way.
In order to check which people a certain node is seeing, a service was implemented inside the node namespace. You can call it by running:
rosservice call /recognition_node/get_seen_faces_names
This service will return a Trigger message response. If nobody is seen at the moment, it returns False and a no face names. Otherwise, it returns True and the names of seen people. If a person is unkown, it will return Unknown in the name field.
Adding new people to the known dataset is currently execute via hardcode, at the setup_detection() function from recognition_node.py. In the future, this will be done via service.
https://github.com/ageitgey/face_recognition
https://github.com/ros-drivers/video_stream_opencv
http://wiki.ros.org/image_transport/Tutorials/SubscribingToImages
