Purdue [CNIT 581-SDR/ Spring 2021] Face Mask Detection Robot

Progress Update Week 13-15: Solved the USB port issue by switching the user to the root Solved the lagging issue by separating the program into a detector and the main program Used UDP communication to connect two programs Sender: Face mask detector Receiver: Main program Calculated the driving control errors  Used Espeak (a text-to-speech generator) as our audio assistant Filmed the demo video Did the project presentation Wrote the final project report Final Product:            Project Demo (If the video cannot be opened, here is the YouTube link: https://youtu.be/JRxGMKsTSpk ) Conclusion The robot is able to check if the passenger wear a mask or not The robot can move towards the passenger without a mask The robot can notify the passenger to take a mask Future work Use a closed-loop system for the driving control or let the robot move randomly Use a smaller computer that can “sit” on the Roomba Detector aborted because moving too fast or hit the camera Improve the mask detector

For hardware construction: A simple camera holder was built to mount the camera with tripod. For software construction: In these two weeks, we tried to embed the Roomba driving control and python audio into our detector code. Because of the port issue, we were not able to connect the Roomba to our original Linux VW. Therefore, we switched to other computers and platforms. The following are the platforms/computers that we tried and the reason why they were not able to complete our project goals. There are three major functions we are looking at:  1. Can it run the mask detecor.py? 2. Can it install/support PyAudio? 3. Can it access Roomba's USB port and control the robot? Table 1 . System configuration and issues Platform Name What it supports What it does not support Why? VMWare Linux VM Mask detector, PyAudio Roomba driving control USB port could not be opened Raspberry Pi Roomba driving control PyAudio, Mask detector "Arm-linux-gnueabihf-gcc" failed with exit status 1;

  What we finished in these two weeks: Hardware for the robots We have used a tripod to hold the mobile camera. Figure 1&2 demonstrate the graphic design of the robot and how it actually looks like. A 10 ft extension USB cable will be used to connect the camera and the computer We estimated the observation distance between the robot and passengers. Between 1.3 meters to 1.5 meters, the robot is able to observe two passengers at once. Figure 1: The hardware design of face mask detection robot Figure 2: How the robot looks like in reality Implementing the face mask detector We applied the face mask detector designed by  Chandrika Deb , which has a 98% accuracy for Face Mask Detection after training via TensorFlow-gpu==2.5.0 The framework used by the detector:  OpenCV Caffe-based face detector Keras TensorFlow MobileNetV2 A short demo video of us testing the face mask detector can be found below. Video 1: Testing the real-time video stream mask detecting Reference:  https://github.com