NVStreamer

NVStreamer is a development tool that provides video files that can be streamed using the RTSP protocol. It can input video streams such as MP4 as RTSP streams instead of an actual camera, which is a convenient function when testing application operation, and you can easily add and delete RTSP streams using the WebUI as shown below.
It can also be run on a Linux workstation running Ubuntu 20.04 or later, so video files stored on the Workstation PC's storage outside the Jetson can be used as an input source.

Video Storage Toolkit(VST)

Video Storage Toolkit (VST) provides video decoding, streaming, and storage functions. It can automatically detect ONVIF-S profile-compliant IP cameras on the local network and use them as video sources. RTSP streams captured by NVStreamer can also be used. In addition, the web-based UI shown below allows for easy video operation and live playback of multiple streams.

VST uses Redis streams to publish status change events of camera devices and to receive metadata when running AI services. Redis is also provided as part of Microservices.

Analytics AI Service

The Analytics AI Service provides functions for detecting the number of people in camera footage (FOV), detecting the number of people crossing designated tripwire lines (Tripwire), and detecting the number of people within a designated area (ROI), and can be used as a behavioral analysis Microservices.
It also has a function to trigger alerts according to user-defined alert rules, which can be used for surveillance camera solutions, etc.

Source: View video overlay & analytics

IoTGateway、Firewall Service

The IoTGateway service enables access to Jetson devices via the cloud, provides authentication and management of access from mobile devices, and provides video streaming between devices.
Additionally, the Firewall service is implemented using the Uncomplicated Firewall (UFW). Please note that this feature is disabled by default.

Monitoring

The Monitoring service provides information about the GPU, GPU and RAM usage, and camera status on the Jetson, and can be used to monitor applications. These features are provided as third-party services (Grafana, Prometheus).

Source: System Metrics

AI Perception Service

NVIDIA DeepStream SDK：

The functionality of the NVIDIA DeepStream SDK, which was previously offered as a separate software package, has also been integrated into Jetson Platform Services. DeepStream can implement functions such as object detection and tracking on input video streams, making it an ideal service for AI image analysis applications.

DeepStream is also explained in the article below, so please take a look if you are interested.

 [Necessary knowledge for AI image analysis application development] Episode 1 What is NVIDIA DeepStream SDK?

Zero Shot Detection：

This AI service is provided as NanoOwl sample software, a pre-built Docker container that can be launched with Docker compose.

Source: NVIDIA Jetson AI Lab

Visual Language Models (VLMs)：

VLM is a multi-modal model that uses large-scale language models and vision transformers that support image and text input. It has a chat function for input images, alert settings based on natural language, and can be linked to a mobile app to display live streams, chat functions, and alert settings from a mobile device.

This AI service is also provided as a pre-built Docker container that can be launched with Docker compose.

The diagram below shows an example of how a mobile app can be used. For example, by setting an alert such as "Is there a fire?" in advance, you can receive a push notification when the live stream of the fire scene is evaluated as True.