Issue link: https://resources.mouser.com/i/1442826
19 AI Au-Zone Technologies Inc. offers many ML demos with tools to optimize the solutions for a wide range of devices. Their DeepView™ Machine Learning Toolkit and RunTime Inference Engine help developers quickly design, train, and deploy DL networks on a wide range of embedded MCUs, CPUs, GPUs, and ML accelerators. Building on the ML Toolkit, Au-Zone's Intelligent IoT Dev Kit provides a selection of proven sensors, processor, and model combinations to support extremely rapid prototyping, customization, and commercial deployment. NXP's eIQ™ machine learning software development environment enables ML algorithms with inference engines, neural network compilers, and optimized libraries specifically targeting NXP MCUs, i.MX RT crossover MCUs, and the i.MX applications processors. This leverages open-source technologies and is fully integrated into the NXP MCUXpresso software development kit (SDK) and Yocto development environments. This takes care of all ML software dependencies, including the necessary hardware abstraction layers that connect the advanced machine learning technology to the underlying compute engines—allowing more manageable development of complete system-level applications. The eIQ software development environment is expanding rapidly. NXP has deployed and optimized technologies, such as CMSIS-NN, TensorFlow, TensorFlow Lite, OpenCV, and Arm NN. eIQ software is accompanied by sample applications in object detection and voice recognition, to provide a starting point in deployment of machine learning at the edge. Further development is ongoing or planned to offer: • Guidelines for importing pre-trained models based on popular NN frameworks, such as TensorFlow, Caffe, and ONNX • End-to-end example applications to demonstrate typical customer use-cases, e.g., camera inference engine • Support for emerging neural net compilers, such as GLOW • A suite of classical ML algorithms, e.g., support vector machine (SVM) and random forest Cloud Solutions Despite the clear benefits of running ML models on local machines, it is impossible to overlook the advantages of popular cloud- based ML solutions from leading service vendors such as Amazon and Microsoft. They offer voice, vision, and other ML services, along with robust development kits that are easy to use, at low cost, and with minimal effort. CHALK TALK: MACHINE LEARNING IS AT THE EDGE Amazon reference solutions can help you create interactive products with Alexa™ Voice Service (AVS) built right in. These kits include chipsets, voice processing technologies, and client software certified to leverage the AVS application programming interfaces running on a wide range of processors. Microsoft is bringing Artificial Intelligence (AI) and Machine Learning (ML) capabilities to Azure IoT users with the Anomaly Detection Solution for Azure IoT, powered by NXP's i.MX RT106C Crossover Processors. This toolbox delivers a full suite of sensors and an associated Anomaly Detection Toolbox, which uses ML algorithms such as Random Forest and Simple Vector Machine (SVM) to model normal behavior of devices and detect anomalous behavior through a combined local and cloud mechanisms. This allows much lower cloud bandwidth requirements while maintaining full online logging and processing capabilities, for applications such as predictive maintenance for rotating components, presence detection, and intrusion detection. Microsoft and NXP have also announced collaboration to deliver a new Microsoft Azure Sphere certified crossover applications processor to run the Azure Sphere security platform seamlessly on an edge node processor that will provide multi-core heterogeneous computing, a rich graphics experience, and low-power audio processing capabilities. NXP continues to partner with leading software and hardware companies, researchers, and distributors to offer ML solutions ranging from generic sample code and reference designs to ready-made, purpose-built, and certified platforms with the flexibility to add your custom features. The technology is ready, and NXP stands prepared and prepared to help develop your machine learning ideas into affordable and secure products for the real world. ■