This article is contributed. See the original author and article here.
We just made building video analytics solutions simpler from edge to cloud with a new Azure IoT Central application template. This application template integrates Azure Live Video analytics video inferencing pipeline and OpenVINO™ AI Inference server by Intel® to build an end to end solution in a few hrs.
The number of IP cameras is projected to reach 1 billion (globally) by 2021. Traditionally, these types of cameras are used for security and surveillance. With the advent of video AI, businesses increasingly want to use their cameras to extract insights that help improve their profitability and automate (or semi-automate) their business processes. Such video analytics applied to live video streams help businesses react to real-time events and derive new business insights by observing trends over time.
Building a video analytics solution involves multiple complicated phases. This is relatively elaborate instrumentation that requires significant technical expertise and time. These solutions typically start with setting up new cameras or leveraging existing IP cameras for video traffic. IP cameras are versatile devices that support comprehensive configuration and management based on ONVIF standards. Once the IP cameras are set up, you need to ingest the video feeds, process the video, and prepare frames for analysis using inference servers that use specific AI models. These inference servers must be highly performant so that the solution can scale to dozens of cameras at any facility. The results from video analytics need to be collected and stored along with the relevant video for business applications to consume.
Using the new Azure IoT Central application template you can design, define, deploy, scale, and manage a live video analytics solution within hours. Video analytics template supports object and motion detection scenarios with key value propositions, as shown in the following illustration.
Figure 1. Customer & Partner value proposition from Video Analytics – Object and Motion detection app template
In our mission to democratize video analytics, Microsoft and Intel collaborated to build end-to-end video analytics solutions using IoT Central. These solutions leverage:
- Live Video Analytics on IoT Edge (LVA) to capture, record, and analyze live video. LVA is a platform for building AI-based video solutions and applications that include AI applications to live video. You can generate real-time business insights from live video streams, process data near the source to minimize latency and bandwidth requirements and apply the AI models of your choice. LVA provides a flexible programming model to design live video workflows and defines an extensibility model for integrating with inference servers. This frees you up to focus development efforts on the business outcome rather than setting up and operating a complex, live video pipeline.
- For real-time analysis of live video feeds, the video pipeline leverages OpenVINO™ Model Server (OVMS), an inference server that’s highly optimized for AI vision workloads and developed for Intel® architectures. OVMS is powered by OpenVINO™ toolkit, a high-performance inference engine optimized for Intel® hardware on the Edge. An extension has been added to OVMS for easy exchange of video frames and inference results between the inference server and LVA, thus empowering you to run any OpenVINO™ toolkit supported model, and select from the wide variety of acceleration mechanisms provided by Intel® hardware. These include CPUs (Atom, Core, Xeon), FPGAs, VPUs.
- Azure IoT Central is a platform for rapidly building enterprise-grade IoT applications on a secure, reliable, and scalable infrastructure. IoT Central simplifies the initial setup of your IoT solution and reduces the management burden, operational costs, and overhead of a typical IoT project. This enables you to apply your resources and unique domain expertise to solving customer needs and creating business value, rather than needing to tackle the mechanics of operating, managing, securing, and scaling a global IoT solution.
The IoT Central application template brings the goodness of Azure IoT Central, Live Video Analytics, and Intel components integration to enable building scalable solutions in a few hrs. as described in tutorials
Figure 2. Block diagram of Video Analytics – Object and Motion Detection app template
The app template stitches the following components,
- Cloud Services – IoT Central Video Analytics Application Template to stich the end-end solution & Azure Media Services for video snippet storage
- Edge Modules – Video processing pipeline (Live Video Analytics), hardware optimized OpenVINO™ AI Inference server by Intel, IoT Central gateway module to for protocol & identify translation of RTSP & Camera, RTSP Server (Live 555) for pre-recorded video strams
- Connecting & managing IP Camera, RTSP streams and AI module configuration
The IoT Central application template natively provides device operators view for object and motion detection scenarios, as shown in the following illustration.
Figure 3. Dashboard from IoT Central template for Video Analytics – Object & Motion Detection
The dashboard in the new Video Analytics – Object & Motion Detection template for IoT Central is shown above. The template requires,
- IP cameras (any IP cameras that support RTSP on the ONVIF conformant products page devices that conform with profiles G, S, or T), or You can leverage simulated video stream that we ship as part of this template for demonstrations.
- Linux server powered by your choice of Intel® acceleration technology (CPUs such as Atom, Core, Xeon, or FPGAs, or VPUs)
- Azure subscription to host relevant cloud services.
Get started today
- You can use the new Video Analytics for Object & Motion Detection template to build and deploy your live video analytics solution.
- You can build Video Analytics solution within hours by leveraging Azure IoT Central, Live Video Analytics, and Intel.
- You can learn more about Live Video Analytics on IoT Edge here, and try out some of the other video analytics scenarios via the quickstarts and tutorials here. These show you how you can leverage open source AI models such as those in the Open Model Zoo repository or YOLOv3, or custom models that you have built, to analyze live video.
- You can learn more about the OpenVINO™ Inference server by Intel® in Azure marketplace and its underlying technologies here. You can access developer kits to learn how to accelerate edge workloads using Intel®-based accelerators CPUs, iGPUs, VPUs and FPGAs. You can select from a wide range of AI Models from Open Model Zoo
Brought to you by Dr. Ware, Microsoft Office 365 Silver Partner, Charleston SC.
Recent Comments