dnn_face_recognition

Recognize faces with YuNet and SFace (cv::FaceRecognizerSF).

Face recognition with YuNet (cv::FaceDetectorYN) and SFace (cv::FaceRecognizerSF), https://github.com/opencv/opencv_zoo/tree/main/models/face_recognition_sface The faces trained with the ~train service are saved to data_dir/<label>/<label>_<time>.png as the aligned face images of SFace, and loaded at the start.

opencv_apps	3.0.0 and later
ROS	ROS One, jazzy, kilted, lyrical and later
OpenCV	OpenCV 4.5.4 or later
Group	Faces, the nodes of which publish `~/faces` (FaceArrayStamped)

Usage

# ROS 2
ros2 launch opencv_apps dnn_face_recognition.launch.xml image:=/camera/image_raw debug_view:=false
# ROS 1
roslaunch opencv_apps dnn_face_recognition.launch image:=/camera/image_raw debug_view:=false

Try it with the demo video

Download the input video of the demo above, publish it as image with image_publisher, and run the node on it, in the terminals:

wget https://raw.githubusercontent.com/ros-perception/opencv_apps/indigo/test/face_detector_withface_test_diamondback.mp4
# terminal 1: publish the video as image in a loop
ros2 run image_publisher image_publisher_node face_detector_withface_test_diamondback.mp4 --ros-args -p publish_rate:=10.0 -r image_raw:=image
# terminal 2: run the node with the parameters of the demo
ros2 launch opencv_apps dnn_face_recognition.launch.xml image:=image data_dir:=/tmp/dnn_face_data
# terminal 3: then run once
ros2 run opencv_apps face_recognition_trainer.py --ros-args -r faces:=/dnn_face_recognition/output -r train:=/dnn_face_recognition/train
# terminal 4: view the result
ros2 run rqt_image_view rqt_image_view /dnn_face_recognition/image

Train a face with the trainer, which asks for the name, then the node recognizes it.

See Credits of the demo videos for the license of the video.

Subscribed topics

Topic	Type
`image`	`sensor_msgs/msg/Image`

Published topics

Topic	Type	Note
`~/faces` (recommended)	opencv_apps/msg/FaceArrayStamped	Recommended, opencv_apps 3.0.1 and later. The same as `~/output`, with the name of the other face nodes.
`~/image`	`sensor_msgs/msg/Image`
`~/output` (compatibility)	opencv_apps/msg/FaceArrayStamped	Kept for compatibility, use `~/faces`.

Services

Service	Type
`~/train`	`opencv_apps/srv/FaceRecognitionTrain`

Parameters

See also the common parameters.

Name	Type	Default (range)	Changed at run time	Description
`backend`	string	“auto”	yes	Backend of cv::dnn: auto, cpu, opencl, opencl_fp16, cuda or cuda_fp16. auto uses CUDA when OpenCV is built with CUDA and a CUDA device is found, and the CPU otherwise.
`data_dir`	string	“~/.ros/opencv_apps/dnn_face_data”	yes	Directory to save and load the trained faces
`detection_model`	string	“”	yes	Path or URL of the YuNet face detection model (onnx), empty for the default for this OpenCV version. The URL is downloaded to ~/.ros/opencv_apps/models/ at the first time
`match_threshold`	double	0.363 (-1 - 1)	yes	Threshold of the cosine similarity to recognize the face as the trained one, 0.363 is suggested by SFace
`queue_size`	int	3	no, restart the node	Specify queue_size of input image subscribers
`recognition_model`	string	“”	yes	Path or URL of the SFace face recognition model (onnx), empty for the default. The URL is downloaded to ~/.ros/opencv_apps/models/ at the first time
`score_threshold`	double	0.9 (0 - 1)	yes	Threshold of the score to detect the faces

Source

src/nodelet/dnn_face_recognition_nodelet.cpp, launch/dnn_face_recognition.launch.xml