This post is a continuation of my tinyML learning journey this summer that is inspired by the Hackster Eyes on the Edge Vision Challenge .


I've been attending as many of the webinars that I can or at least watching the recordings on YouTube.


One that I found particularly interesting last month was the "tinyML Vision Challenge - Intel-Luxonis Vision Platform" video


I've followed Luxonis for a while because of their use of the Intel Movidius Myriad X Vision Processing Unit (VPU).  I first encountered the Myriad X in my Tello Drone Tello Drone Intro Blog and later when using the Intel Neural Compute Stick (NCS) on a PYNQ-Z2 Intel Movidius NCS on PYNQ-Z2 .  Luxonis makes some interesting Spatial (stereo) cameras that utilize their DepthAI platform.  I came close to getting an OAK-D (OpenCV AI Kit) camera , but never felt I had the time to invest in integrating it with a Raspberry Pi4 and also struggled a bit using the OpenVINO platform for programming the Myriad X.


There have been two developments since I last used OpenVINO with the NCS that caught my attention:

  1. OpenVINO (Open Visual Inference and Neural network Optimization) now has a fully developed cloud based framework that supports all of Intel's hardware AI platforms.
  2. Luxonis has developed a new Spatial AI camera board that integrates an ESP32 that facilitates developing a portable battery powered Edge AI solution. LUX-ESP32 (OAK-D-IoT-40).




Luxonis LUX-ESP32



The devcloud based framework removes the complexity of installing and maintaining a local OpenVINO setup that I have had problems keeping up-to-date..  And the LUX-ESP32 seems to be perfect for experimenting with Spatial AI.  I was fortunate that as part of the Eyes on the Edge Vision Challenge, I was able get a 50% discount on a LUX_ESP32.  I'm looking forward to seeing what I can do with OpenVINO and Spatial-AI .