Testing

Initial Testing Plan

After we get hardware running, we will perform a multitude of tests to figure out the optimal configuration for our system. The initial configuration we are continuing from consists of a Raspberry Pi running Frigate, and a PyTorch YOLO derived neural network. Since YOLO was used for the neural network, a Google Coral AI accelerator could not have been used, since YOLO is currently lacking good support for the TensorFlow software library, which is what Coral uses (and Frigate strongly recommends in their documentation). 

Testing will proceed by using the previous model as a control, and replacing it with a TensorFlow based Neural Network that the Coral accelerator accepts. Testing will then proceed to pit Pi + Coral accelerator against a Coral Dev Board Micro, in not just performance, but power consumption performance as well. Once we evaluate the best course of action, we will focus on tuning the models to better deal with ear detection.