Face Recognition Ultra Light

The light-weight ROS face recognition package based on Python3. A module of SMARTBat.

Installation

• Install ROS melodic
• astra_camera, we used stereo_s_u3 by default
• Install python packages, you might need to upgrade your pip version at first
  • pip3 install -r requirements.txt

Emotion Detection over mqtt

• run roslaunch face_recognition_ultra_light face_recognition_ultra_light.launch
• rosrun face_recognition_ultra_light mqtt_emotion_subscriber.py to receive emotions from the publisher
  • or run python3 mqtt_emotion_subscriber.py located in face_recognition_ultra_light/scripts
  • we use mqtt not ros since the special situation under COVID-19 quarantine

Face Training

• To add training data, you can use faces/recorder.py

  cd faces
  python3 recorder.py

  To quit the program, press q
• After having a video for training, you should move it to faces/training/<name>/<name>.avi
• Then run python3 training.py to create your model
• Finally you can run the launch file

  roslaunch face_recognition_ultra_light face_recognition_ultra_light.launch

RQt Graph

Result

Reference

• https://towardsdatascience.com/real-time-face-recognition-with-cpu-983d35cc3ec5
• https://mc.ai/emotion-recognition-using-keras/
• Chen, Sheng, et al. "Mobilefacenets: Efficient cnns for accurate real-time face verification on mobile devices." Chinese Conference on Biometric Recognition. Springer, Cham, 2018.
• Simonyan, Karen, and Andrew Zisserman. "Very deep convolutional networks for large-scale image recognition." arXiv preprint arXiv:1409.1556 (2014).