NVIDIA and Amazon Web Services bring AI to IoT devices

NVIDIA today announced a collaboration with Amazon Web Services (AWS) IoT on NVIDIA Jetson to enable customers to deploy artificial intelligence (AI) and deep learning to millions of connected devices.

The joint solution enables models to be easily created, trained and optimised on AWS, then deployed to Jetson-powered edge devices using AWS IoT Greengrass. The NVIDIA Jetson platform offers AI at the edge with high-performance, power-efficient computing. Customers can save bandwidth and cost by running near real-time inference directly on a Jetson-powered device instead of sending data to the cloud first. 

AWS IoT Greengrass seamlessly extends AWS to edge devices, including offering machine learning inference, so they can act locally on the data they generate while still using the cloud for management, analytics and long-term storage. The data is then sent back to machine learning services such as Amazon SageMaker to improve model accuracy. 

Jetson comes with software tools and software development kits (SDKs), including NVIDIA JetPack. It also supports multiple frameworks such as MXNet, Caffe, TensorFlow, and PyTorch, so developers can use these algorithms to quickly deploy real-world applications. 

“Jetson is a high-performance computer that’s built on the same architecture and unified software that powers the world’s fastest supercomputers,” said Deepu Talla, VP and GM, Autonomous Machines at NVIDIA. 

“We offer the new Jetson Nano all the way to Jetson AGX Xavier, so IoT devices can now scale from small IoT devices to powerful IoT gateways.” 

Applications include autonomous machines and smart cameras for industries such as retail, manufacturing, agriculture, and more. In high-precision agriculture, Jetson-powered cameras running AWS IoT Greengrass can target weeds in near real time, capture previously-unidentified weeds, upload anomalies to the cloud, and retrain and deploy the model quickly. 

With edge-powered automated optical inspection, product defects can be quickly identified on factory floors. In retail, Jetson-powered devices can monitor inventory, customer behaviour, as well as checkout and process data on the edge, using AWS IoT Greengrass to deploy trained neural networks.

Hashtag: #GTC19

*NVIDIA sponsored transport and accommodation for GTC.