This week, I worked on doing research on the computer vision aspect of our project. Specifically we are determining which approach of facial detection we want to implement. We have successfully narrowed it down between using the Dlib algorithm for face detection, facial landmarking and eye detection or using OpenCV’s DNN algorithm. Both algorithm have its sets of trade offs including performance and accuracy. We are currently learning towards implementing the OpenCV’s DNN algorithm for facial detection because it has a higher frames per second rate and a better accuracy. However, it is an extensive algorithm that may or may not be able to run on our boards GPU. The final decision will be based on the compatibility of the board and the polarizability of the algorithms. As a backup, we will utilize the slower and less accurate Dlib algorithm for face detection since that is able to run on our boards GPU.
Lastly, I worked with my team to finish our proposal presentation slides.