Closed mitchross closed 3 years ago
Maybe. I am not using the edgetpu library wrapper because it's slower. I am doing a lot of the prep work to resize and convert the image into a numpy matrix with the right dimensions in an entirely separate process. The detection process does nothing other than grab a presized/preformatted input from memory, run detection, and hand off the output. Since I am splitting the work across multiple processes, these wrapper libraries don't typically fit my use case.
Maybe. I am not using the edgetpu library wrapper because it's slower. I am doing a lot of the prep work to resize and convert the image into a numpy matrix with the right dimensions in an entirely separate process. The detection process does nothing other than grab a presized/preformatted input from memory, run detection, and hand off the output. Since I am splitting the work across multiple processes, these wrapper libraries don't typically fit my use case.
Thanks for explanation!
edgetpu from coral is deprecated, you can use the PyCoral library to reduce complexity of your edgetpu.py code
See -> https://github.com/google-coral/pycoral/blob/master/examples/detect_image.py
API -> https://coral.ai/docs/reference/py/