I did not use the edgetpu. There's a way to work that but I decided to not go for that direction since there are less support. I use pycoral package -- it means that coco object detection code is not going to work since it still use edgetpu. Two ways to solve this: 1. Modify code yourself to adept it to the pycoral api 2. Use a different test code file.

I chose 2nd solution since it is easier and we only need to test object detection to see whether it works or not. I use the pycoral's examples to test it. The camera object detection file could be found from the pycoral page, and I put the link here:

https://github.com/google-coral/examples-camera

Choose a way you like and follow instructions to test it. I use the opencv one.

For next few sections I believe there might need actual code modifications to fit for pycoral -- but I haven't proceed to that yet.