Edge AI
Lowest latency, power and cost for multi-sensor analytics and machine learning applications at the Industrial IoT edge
Analytics and Machine Learning encompass a tremendous field of industrial applications, for instance Predictive Maintenance, Digital Twin model based control, anomaly detection and many other use cases. AMD and the AMD ecosystem offer multiple different approaches to address these Edge applications based on user trends.
Analytics and Machine Learning
Edge AI Platform
Over the course of the past four years, extensive work has been done by AMD to develop a complete end to end flow which allows software developers, hardware developers and data scientists to leverage the existing AI/ML ecosystem. In this paradigm, we have designed tools (Vitis AI) which enable our customers to directly parse the model graph and trained weights that are saved out of popular ML frameworks. Today, this includes Caffe, Pytorch and TensorFlow. We have developed pruning, quantization tools, compiler, runtime and efficient programmable IP which allows us to deploy your network on a variety of platforms, but on the edge or in popular cloud and server architectures.
PYNQ - Python on Zynq
Python powered control, edge analytics and machine learning enabled by PYNQ. PYNQ is a software-hardware framework for AMD adaptive SoCs leveraging the programmable hardware to pre-process sensor and other types of data to make software analysis and manipulation highly efficient in an embedded processor. PYNQ supports all major python libraries like Numpy, Scikit-Learn, and Pandas etc.
Cloud Provider
The trend in Industrial is a partial shift of processing from the Cloud to the Edge driven by:
- Physical assets demand low latency decisions/actions closest to the data acquisition point (typically under 10 ms)
- Data generated is typically large in size and moving and storing all generated data is not desirable due to OPEX cost, time and privacy concerns
AMD provides the industry’s most capable single-chip Edge embedded processing platforms to address such trends. Furthermore, AMD adaptive SoC portfolio and ecosystem partnerships with leading cloud service providers enable distribution of tasks across the Cloud and Edge as well as mobilize applications from the Cloud to Edge.
AWS IoT
Industrial IoT is also rapidly accelerating the opportunity for cloud connected and collaborative control systems that can unlock the next set of capabilities of the industrial asset using machine learning. Industrial control system providers are realizing this vision and the need for integrated edge to cloud solutions that will accelerate their time to market. AMD with AWS IoT provides differentiated and collaborative edge-to-cloud machine learning capabilities.
- AWS IoT Greengrass: Seamlessly extends AWS to edge devices so they can act locally on the data they generate, while still using the cloud for management, analytics, and durable storage
- AWS FreeRTOS: Operating system for microcontrollers that makes small, low-power edge devices easy to program, deploy, secure, connect, and manage
- AWS Sagemaker: Fully-managed service that covers the entire machine learning workflow to label and prepare your data, choose an algorithm, train the algorithm, tune and optimize it for deployment, make predictions, and take action
Microsoft Azure IoT
Industrial IoT is also rapidly accelerating the opportunity for cloud connected and collaborative control systems that can unlock the next set of capabilities of the industrial asset using machine learning. Industrial control system providers are realizing this vision and the need for integrated edge to cloud solutions that will accelerate their time to market. AMD with Microsoft Azure IoT provide differentiated and collaborative edge-to-cloud machine learning capabilities.
- Azure IoT Hub: Connect, monitor and manage billions of IoT assets
- Azure IoT Edge: Extend cloud intelligence and analytics to the edge devices
- Azure IoT Digital Twins: Build next-generation IoT spatial intelligence Solutions
SoC Integration
Solutions
| Solution Provider | Description | Device Support |
|---|---|---|
| AMD | Kria Vision AI Starter Kit – KV260 | AMD Kria K26 SOM |
| AMD Kria Ecosystem | Kria App Store | AMD Kria K26 SOM |
| AMD | Why AMD AI? | |
| AMD - Edge AI Platform | Vitis AI Edge Edge White Paper |
AMD adaptive SoCs Kria K26 SOM Versal AI Edge |
| AMD - PYNQ | PYNQ Homepage PYNQ Community Projects |
AMD adaptive SoCs Kria K26 SOM Versal AI Edge |
| AWS IoT | AWS Certified AMD Products AWS IoT AMD – AWS Workshop |
AMD adaptive SoCs Kria K26 SOM Versal AI Edge |
| Azure IoT | Azure IoT | AMD adaptive SoCs Kria K26 SOM Versal AI Edge |
| AMD Tools | Vitis software platform Vivado ML |
AMD adaptive SoCs Kria K26 SOM Versal AI Edge |
| AMD | SPYN Design Files |
AMD adaptive SoCs Kria K26 SOM Versal AI Edge |
Some Industrial and Healthcare IoT products need all elements of the AMD IIoT and HcIoT Solutions Stack, all need some. The AMD IIoT and HcIoT Solutions Stack is comprised of optimized AMD and Ecosystem building blocks and solutions used across Industrial and Healthcare IoT platforms. Starting from scratch is never something you will have to do with a AMD-based Industrial or Healthcare IoT system. Minimize development time and cost and maximize design reuse on your next Industrial or Healthcare IoT platform by exploring the different elements of the AMD IIoT and HcIoT Solutions Stack.
