This week I was able to have the set up for OpenCV up and running on the Jetson TX2. After doing research and looking into haar cascade models that are available to the public, we were able to get it working for facial recognition. I think that this is most likely the option to go with but more testing needs to be done on how well it can detect a face. I’ve also optimized the program to be less laggy by incorporating the timing of the facial detection as it was trying to detect faces on every single frame that it was able to get. The amount of processing that has to be done is too much to be completed in a single frame which led to a lag of about 1 seconds. By reducing the frequency of processing, I was able to achieve a facial tracking that is a little laggy but it no longer lags the actual video feed. With the detection working, I’ve also made a rough estimation of how we can attribute location in the camera feed itself into an angle for the servo to move in. Now we have begun looking into threading to actually allow for our separate components of the project to communicate with each other. I’ve also considered looking into a Go library for python so we can make use of the concurrency model that exists in Go.

I’ve also begun planning out how to make use of this simple detection model to detect actual speaking people as well since that is the main goal here. My first thoughts for speaker detection is to first detect the lips of a face and try to take the location and measurements over multiple frames to see if I can correlate it to speaking. Outside of individual work I also worked on the design review presentation.

For the upcoming week the main goals are to finalize the output format for communications and also help Heather work with multiple camera setup with CV since that is quite important. We also have to work on the Design Review Paper which will probably take a long time.