sayter99 / face_recognition_ultra_light Goto Github PK

Create a window filter for a persons emotion that have a variable t that that is how far back we check in time to average

I've uploaded mqtt programs (using paho-mqtt) in scripts folder for reference.
TODO:

1. adapt our code to use mqtt
   • our final output will be a single string containing
     • time (not ros time anymore, can use python datetime)
     • name
     • emotion or emotions (avg one)
2. modify the program to publish the message to mqtt broker (topic name TBD)

Create a launch file based on ros_astra_camera/launch/stereo_s_u3.launch. You could copy this file to our repo as a starting point. Since we will modify some values in this launch file, I think it's good to put all details in our launch file.

Thus, the structure would look like

# stereo_s_u3 content
# note: set these two args as true to enable pointcloud2 for testing the face distance
# for our final version, we will turn off it because visualization is not important
<arg name="depth_processing"                default="true" />
<arg name="depth_registered_processing"     default="true" />
# our three nodes (output to screen)

To verify it works, after launching, you can open rviz and see a pointcloud2 topic and if face detected (not unknown), you can see a tf frame of that face.

as title