How to enable inference acceleration on ESP32S3

Question/Issue:
I’ve notice that there are some code under the porting folder of edge impulse sdk for accelerating inference on esp module by using ESP_NN. How should I enable them?

I deploy the inference model by using the Arduino library option. Then how should I explicitly enable the acceleration for ESP32S3?

Hello, @xiqwertyui !
There is no need to explicitly enable ESP NN when using Arduino library deployment option. It is automatically enabled for regular ESP32 and ESP32-S3.

How to disable the ESP NN when using Arduino library deployment option? I am trying to compare the performance of ESP32-S3 with and without ESP-NN library

Best regards

The README file for the ESP_NN library appears to mention how to disable it, and the official documentation provides some test results.

Judging by the given results, enabling acceleration can speed up inference by ten times or more; there is no reason not to use it.

1 Like