I am working with an image impulse that I want to put on the Portenta, Himax WEI and Nano33Ble (with the OV7670 cam)
Can we train our impulse and then change the Latency calculations for different boards, before building the product, or do we have to set the Latency calculations before training?
I am not sure, but you can always alter the dev board to calculate the latency before or after model training process. The latency calculated are most probably MCU centric and do not interfere in the results in the model trained.Since They are an indicative display for the calculations performed on the MCU as a function of time and so, most probably you can view the latency pre or post build
Make sure to correct me if I am wrong
@Rocksetta I am not sure but I guess its probably used for doing the calculations while deploying the model using EON
@Rocksetta they are altered dynamically, so you can change these as you go, no need to retrain.
Current active one is shown when you hover: