How to deploy the Tensorflow weights of a image classification CNN on Arduino

Hi,

I was wondering if there was a tutorial for how to deploy the weights of a Tensorflow model onto an Arduino ? If not, does anyone have any starting point suggestions ?

I'd like to do all the training on a PC, and then deploy the weights of an image classification CNN onto a Forward Pass Network on the Arduino.

The Arduino will need a camera input attached to it, and be processing real-time images, so that the classifier can sound an alert if it spots something it recognises.

Also the Arduino will need a speaker to sound an alert, or play back some audio if the classified image appears infront of the camera.

T

Processing real-time images?
Seriously?
You don't know much about the Arduino do you?

The Arduino Uno has 2048 bytes of RAM. You should consider something like the Raspberry Pi instead, which is very inexpensive yet runs Linux. Several different camera modules are available that interface directly to the RPi board.