I’m currently working with a Grove Vision AI V2 board, equipped with an OV5647 camera, and the XIAO ESP32S3. I’ve managed to successfully upload a custom trained model (using Edge Impulse) onto the Grove Vision AI V2. However, I’ve encountered an issue with enabling the boot mode despite trying multiple boards, operating systems, and type-C cables.
I’ve discovered that using Sense Craft AI and a somewhat complex procedure, it is possible to flash a model and utilize the XIAO along with the “Seeed_Arduino_SSCMA” library to receive images and the model output through Arduino.
My question is: Is it feasible to dynamically flash different models using Arduino? For instance, if I am currently running a Cat detection model but suddenly decide to switch to a Dog detection model (assuming both models are small enough to fit into the XIAO memory), can I flash the Grove Vision V2 without having to go through Sense Craft AI?
As far as I know, it is feasible to dynamically flash different models using Arduino on the Grove Vision AI V2 board. You can leverage the “Seeed_Arduino_SSCMA” library to receive images and model output through Arduino, as you mentioned.
Hi @liaifat85,
Thank you very much for your answer.
Do you, or anybody who is reading this, have an idea on how to do that?
Unfortunately the library is not documented very well, and I guess only .uf2 files can be flashed into the device. I don’t think the write() method is built to flash the board.
Hi there,
I know there are examples and folks will be quick to point those out, However this is probably the best one I have used and it gives great results when using the a smaller transfer learning model.
Check this guide and see if it helps.
edge has very good documentation, Seeed is trying to catch up…LOL
HTH
GL PJ
You can take a look at this Wiki, which does a good job of talking about how to flash the model on the Arduino development board so as to use the Grove vision ai v2
And there are some examples of ready-made models in the library example.
PJ 
