Detect faces with sample_face_detection - Qualcomm Dragonwing Documentation

The models have the source from the Qualcomm AI-hub. It can detect face and locate facial features from a face image. It captures the face_image.jpg as input and publishes the result to the /mediaface_det_image topic.

For model information, see MediaPipe-Face-Detection - Qualcomm AI Hub.

The following figure shows an example face detection result.

Face detection result with a bounding box and a green facial landmark mesh overlaid on a person in a live camera view.

`sample_face_detection` pipeline flow

The following figure shows the pipeline flow for sample_face_detection.

Face detection pipeline from image capture through the face detector, neural network inference, and landmark postprocessing to the result topic. — sample_face_detection pipeline.

ROS nodes used in the `sample_face_detection` pipeline

The following table lists the ROS nodes used in the sample_face_detection pipeline.

ROS node	Description
`qrb_ros_face_detector`	`qrb_ros_face_detector` is a Python-based ROS jazzy package that processes face images. This ROS node subscribes to an image topic and publishes the face image result topic after preprocessing and postprocessing.
qrb_ros_nn_inference	`qrb_ros_nn_inference` is a ROS2 package for performing neural network model, providing AI-based perception for robotics applications.
qrb ros camera	Qualcomm ROS 2 package that captures images with parameters and publishes them to ROS topics.
image_publisher_node	`image_publisher` is a ROS 2 jazzy package that publishes the image ROS topic with a local path.

ROS topics used in the `sample_face_detection` pipeline

The following table lists the ROS topics used in the sample_face_detection pipeline.

ROS topic	Type	Published by
`/mediaface_det_image`	`< sensor_msgs.msg.Image >`	`qrb_ros_face_detector`
`/image_raw`	`<sensor_msgs.msg.Image>`	`image_publisher_node`, `camera_node`
`/face_detector_input_tensor`	`<qrb_ros_tensor_list_msgs.msg.TensorList>`	`qrb_ros_face_detector`
`/face_detector_output_tensor`	`<qrb_ros_tensor_list_msgs.msg.TensorList>`	`qrb_ros_nn_inference`
`/face_landmark_input_tensor`	`<qrb_ros_tensor_list_msgs.msg.TensorList>`	`qrb_ros_face_detector`
`/face_landmark_output_tensor`	`<qrb_ros_tensor_list_msgs.msg.TensorList>`	`qrb_ros_nn_inference`

Prerequisites

You have set up the device, installed ROS2 Jazzy and Qualcomm Intelligent Robotics (QIR) SDK on the device according to Install the QIR SDK.

Build from source of face detection

Build the face detection sample from source when you need to change the sample code. The device steps install the dependencies, clone the source repository, and build the package.

Device steps

Download the sample face detection model

Download the models

mkdir -p /opt/model && cd /opt/model
sudo wget https://huggingface.co/qualcomm/MediaPipe-Face-Detection/resolve/0dd669a326ec24a884e51b82741997299d937705/MediaPipeFaceDetector.bin -O MediaPipeFaceDetector.bin
sudo wget https://huggingface.co/qualcomm/MediaPipe-Face-Detection/resolve/0dd669a326ec24a884e51b82741997299d937705/MediaPipeFaceLandmarkDetector.bin -O MediaPipeFaceLandmarkDetector.bin
sudo wget https://raw.githubusercontent.com/zmurez/MediaPipePyTorch/65f2549ba35cd61dfd29f402f6c21882a32fabb1/anchors_face.npy -O anchors_face.npy

Install the dependencies

sudo apt install ros-dev-tools
sudo apt install ros-jazzy-qrb-ros-camera

Download source code from the qrb_ros_sample repository and build

Download and build

mkdir -p ~/qrb_ros_ws/src && cd ~/qrb_ros_ws/src
git clone https://github.com/qualcomm-qrb-ros/qrb_ros_samples.git
cd ai_vision/sample_face_detection
rosdep install -i --from-path src --rosdistro jazzy -y
colcon build

Run the sample_face_detection sample application

Run the sample application

source install/setup.bash
# You can use the default face image file
ros2 launch sample_face_detection launch_with_image_publisher.py model_path:=/opt/model/
# You can also replace this with a custom image file
ros2 launch sample_face_detection launch_with_image_publisher.py image_path:=/opt/resource/xxx.jpg model_path:=/opt/model/
# You can launch with qrb ros camera
ros2 launch sample_face_detection launch_with_qrb_ros_camera.py  model_path:=/opt/model/

​sample_face_detection pipeline flow

​ROS nodes used in the sample_face_detection pipeline

​ROS topics used in the sample_face_detection pipeline

​Prerequisites

​Build from source of face detection

​Device steps

`sample_face_detection` pipeline flow

ROS nodes used in the `sample_face_detection` pipeline

ROS topics used in the `sample_face_detection` pipeline

Prerequisites

Build from source of face detection

Device steps