Announcing Official Support for the Espressif ESP-EYE (ESP32)

Blog
#1

We are thrilled to announce the official support for the Espressif Systems ESP32 system-on-chip (SoC) and the ESP32-based ESP-EYE development board. 

This is a companion discussion topic for the original entry at https://www.edgeimpulse.com/blog/announcing-official-support-for-the-espressif-esp-eye-esp32
2 Likes
#2

Would love to see projects on ESP32-S3 based boards running! (with esp-nn optimisations)!

FOMO model mentioned in this article should definitely run around 100ms per inference!

1 Like
#3

Indeed! I’d be curious to see S3 vs base ESP32 performance when using Edge Impulse.

@mathijs
Are Edge Impulse exported projects already able to take advantage of S3’s SIMD instruction sets/NN optimizations? Do you have any results?

#4

Great - I tried it out with an existing project Goodwatch - Voice controlled alarm clock that resulted in half of the classificaiton time.

"Old:" Predictions (DSP: 87 ms., Classification: 21 ms., Anomaly: 0 ms.): 
"ESP-NN:" Predictions (DSP: 77 ms., Classification: 11 ms., Anomaly: 0 ms.):

I am using Platformio and Arduino as a component - to get this working with the new EI firmware I needed to do:

  • replace edge-impulse-sdk with the new firmware.
  • add esp-nn (espressif) to my project
  • fix missing path in conv.cc and fully_connected.cc by adding “edge-impulse-sdk/”
  • Enable ESP-NN via sdkconfig.

Source code (maybe good as example) on GitHub

!! Cant wait for next release - to get the DSP part addressed :slight_smile:

1 Like