our question is to understand the hardware requirement to run ML program,
suppose if we configure the ML model in the Edge impulse platform as explained by them, and then download in our custom nrf52840 board, whether to run the program do we need another board to be attached to our custom hardware or we can run with out it.
whether we have to buy and attach a different board for interfacing with our nrf52840 custom board to run the program or we can run directly in our custom board,
we understood that the configuration can be done on the impulse, configuration of the sensors as well as the ML model can be done on the impulse platform as explained on their video and same can be downloaded in our custom board but our question is weather for this program to run do we need a different board to be attach to run with our board or directly we can run it
Hi @rajender, the model that comes out of Edge Impulse runs directly on the nRF52840, no extra board required. Think of it as a peripheral driver, but rather than raw sensor values you get conclusions out of it.
Hi janjongboom,
is that mean i can tryout edge impulse final outputs without “X-NUCLEO-IKS02A1 board” using nrf52840DK board
Correct, you just get a library out and can hook any sensor into that. E.g. see an example of classifying data from any accelerometer (whether its on a shield or just mounted on the PCB) here; https://docs.edgeimpulse.com/docs/cli-data-forwarder#example-zephyr
Here’s a really simple example that uses the on-chip (die) temperature sensor: https://github.com/edgeyoda/example-standalone-inferencing-zephyr-nRF-internal-temp-sensor
No external hardware required. This might be useful for you to see how things are setup in Zephyr.
1 Like