As AVNet Silica partners, Deep Vision Consulting had the chance to test the Neural Processing Unit (NPU) of the i.MX 8M PLUS before the official launch which has taken place on March 2021 at Embedded World.
The Neural Processing Unit (NPU) of the i.MX 8M PLUS unleashes great computational power for running live inference of deep learning on the edge.
What you see in this video is a use case we made up just as showcase: it’s a resnet-like network taking as input images and weights (from the scale), treating them jointly to produce fruit classification.
Is it possible to run the inference you see here on a fanless embedded system without creating problem to the rest of the software? Yes and no.
If you want the quick answer, just check this
but if you want to dig the details of it, watch the 20 minutes seminar here
Don’t miss the last part on unconventional networks (object detection on 5Megapixel images or inference on 3D point clouds)