In this category we'll be investigating all aspects of the Artificial Intelligence Edge and its associated hardware. We'll be reviewing the latest hardware and algorithms, as well as conducting investigative hardware projects. We'll be answering, ideally, the right questions right at the cutting edge of Edge AI including: What embedded system hardware is best suited to a particular deep neural network task? What Edge AI hardware would be better suited for use on a drone, robot, etc.
Interlude: TensorFlow Models on the Edge
- Written by BP BP
- Category: Edge AI: Machine Learning Classification Edge AI: Machine Learning Classification
- Published: 22 November 2021 22 November 2021
In this short interval, in the basic classification series, I discuss the role of the Tensorflow Delegate and demonstrate how a compiler is used to generate the components used by hardware accelerators in general and specifically the Coral EdgeTPU.
Machine Learning Basic Classification (3/4): On TI’s EdgeAI Cloud Tool
- Written by BP BP
- Category: Edge AI: Machine Learning Classification Edge AI: Machine Learning Classification
- Published: 13 November 2021 13 November 2021
In this article the TFLite model, developed previously, is used to perform inferencing on TI’s Edge Cloud server and like before predictions are made, using the TFLite interpreter runtime engine.
Machine Learning Basic Classification (1/4): On the Mac Mini M1
- Written by BP BP
- Category: Edge AI: Machine Learning Classification Edge AI: Machine Learning Classification
- Published: 03 November 2021 03 November 2021
In this article I use the Mac Mini M1 with the Tensorflow-metal plugin to classify clothing on the M1's GPU. Part 1 of the SK-TDA4VM Machine Learning Workflow, using the Tensorflow deep learning framework with the Fashion MNIST dataset.
Machine Learning Basic Classification (2/4): TensorFlow Lite Model Conversion
- Written by BP BP
- Category: Edge AI: Machine Learning Classification Edge AI: Machine Learning Classification
- Published: 08 November 2021 08 November 2021
In this article I convert the Fashion MNIST model, created previously, into a TFLite model that can be used to delegate machine learning tasks to the SK-TDA4VM Starter Kit’s hardware accelerators.