Hi folks,
Based on another discussion it became clear that it would be very useful *if Tensorflow.js would support hardware acceleration on Coral TPU devices via USB.
Because then we could make professional AI accessible in Node-RED to users that are less technically skilled. Indeed since Tensorflow.js is javascript you can install it very easy and run it on NodeJs, like we do already with Node-RED. And no need to start worrying about Python and other technical stuff.
For example to execute object detection on camera images, you simply would have to install the node-red-contrib-tfjs-coco-ssd node (from our friend @dceejay) and plug an affordable 55$ Coral TPU stick into the USB3 port of a Raspberry PI 4:
And then you would have hardware acceleration running, since Dave uses Tensorflow.js in his node already ...
Since Tensorflow.js does NOT support the USB devices yet, I asked them what their planning is. They have responded kindly and very fast and seem to be willing to start working on this! However they need some support from an external community like ours:
If we had a SIG subgroup that focuses on the development of this sort of feature, who is interested in being part of that working group so we can get an idea of who may be able to help lead that effort from the external community too.
Unfortunately I neither have the time nor the knowledge to do this. So it would be nice if anybody else could volunteer for this. Because it would be a huge help for us to build a professional video surveillance system with Node-RED!
Of course when something needs to be developed, you can always count on me (and others) to assist!
Thanks a lot !!
Bart
