BT

Facilitating the Spread of Knowledge and Innovation in Professional Software Development

Write for InfoQ

Topics

Choose your language

InfoQ Homepage News NVIDIA Introduces Metropolis Microservices for Jetson to Run AI Apps at the Edge

NVIDIA Introduces Metropolis Microservices for Jetson to Run AI Apps at the Edge

This item in japanese

NVIDIA has expanded its Nvidia Metropolis Microservices Cloud-based AI solution to run on the NVIDIA Jetson embedded platform, including support for video streaming and AI-based perception.

According to Nvidia, its Nvidia Metropolis Microservices solution can accelerate the creation of cloud-native, flexible, and secure vision AI applications running at the edge leveraging its NVIDIA Jetson platform, thanks to new APIs and microservices.

These new Jetson microservices empower developers to modernize their AI application stack, streamline processes, and safeguard applications for the future.

Metropolis Microservices are packaged as a set of Docker containers, which should make it possible to build modular and easily deployable systems. Developers can choose whichever microservice or component they need for their solutions and disregard the rest, NVIDIA says.

The new extensions for Jetson support video storage and media management, tracking algorithms, system monitoring, and IoT services for edge-to-cloud connectivity, as well as microservices for login, authentication, secure access to device APIs, and analytics.

The Video Storage Toolkit provides hardware-accelerated video decoding and encoding, with support for H264 and H265 streaming using standard protocols like RTSP and WebRTC. The Analytics Service enables tracking real-time movement of various objects in live or recorded video. AI perception is based on NVIDIA’s DeepStream SDK, a streaming analytics toolkit for multi-sensor processing, video, audio, and image understanding.

Additionally, NVIDIA created two pre-built reference applications, an AI-powered Network Video Recorder (AI-NVR) and a zero-shot detection using generative AI, to showcase its solution.

The AI-NVR reference workflow integrates all available microservices and provides an example of how to build a complete pipeline including video ingestion, AI inference, analytics, and monitoring. NVIDIA is also providing a client app for Android to demonstrate the use of these APIs on the client side.

The zero-shot detection reference application, on the other hand, shows how to create generative AI applications running at the edge. The application can be controlled remotely over a REST API to detect objects belonging to the category specified in the prompt.

NVIDIA Jetson is a platform for Edge AI and robotics apps able to run up to 275 trillion operations per second (TOPS), says NVIDIA. It supports many open-source models that have been specifically optimized for it and that you can find ini the Jetson Generative AI Labs.

About the Author

Rate this Article

Adoption
Style

BT