Thanks for the pointers to the NCS2 workaround. However if you check out the main analysis section you’ll see I have benchmark numbers for the Coral hardware on the Raspberry Pi 4.

See https://blog.hackster.io/benchmarking-machine-learning-on-the-new-raspberry-pi-4-model-b-88db9304ce4 for the initial benchmarks for the Coral hardware and methodology.