3 minute read
Through our recent acquisition of Scailable, we are greatly enhancing the ease with which developers creating advanced video solutions using the Nx Enterprise Video Operating System (and Nx Toolkit) can add highly performant edge AI functionality to their applications. In this blog post, we will introduce the latest addition to the Nx Toolkit, Nx AI Manager, set to debut in a private Beta alongside the upcoming Gen 6 release.
Why Add AI to Your Nx Video Solution?
Before digging into the Nx AI manager, it is worthwhile to first think about the world of opportunities in front of us when coupling video with highly performant edge AI technology. We live in an era in which AI is rapidly advancing and, within the last few years, we have seen the emergence of small, highly accurate object detection, classification, and image segmentation models. Training AI models for specific vision tasks has become increasingly accessible, to the extent that platforms like Teachable Machine make it child's play to create a proof of concept for image detection.
The ease by which special-purpose AI models can be created to work on top of a video stream opens up endless possibilities. Of course, we are all familiar with solutions in the security industry (i.e., recognizing intruders) and in traffic (i.e., recognizing license plates). But, what about a video + AI solution that alerts a restaurant manager when a table has been left unattended for too long? Or that a glass is empty? Or how about a video + AI solution that can distinguish between weeds and crops, optimizing precision during harvesting processes? These solutions become simple once video streams are effectively managed on a fleet of edge devices (a feature long provided by the Nx platform) and AI seamlessly integrates with the video streams (precisely the enhancement brought by the Nx AI manager).
Making Edge AI Easy and Scalable
The Nx AI manager is a plugin specifically designed to add AI functionality – on the edge – to video solutions built using the Nx Enterprise Video Platform. Via Nx Cloud or the Nx client, the Nx AI manager allows you to configure an AI pipeline, add one or more AI models for each channel, and efficiently use the AI model output in your larger application. And, on top of all that, the Nx AI manager allows you to effortlessly move between hardware targets such that your solution can be built on the best hardware for the job.
Configuring AI Pipelines
An AI pipeline essentially encompasses the entire process from inception to completion, starting with data input from a sensor (such as a camera in vision applications), leveraging AI technology, and ending in visualization, alerts, or metadata logging. The Nx AI manager makes the process of creating an edge AI pipeline a breeze:
- To begin, simply select an Nx channel as input for the AI model. No need to resize, recolor, or otherwise write challenging (and often hardware-dependent) code to decode the channel and ensure images are fed into the AI model.
- Next, configure an AI model; the AI model can be thought of as a function that takes an image as input and provides a class, bounding-box(es), or segment(s) as output. These functions are becoming easier and easier to create (and many reference architectures, such as YOLO, are easily available). As long as these can be stored in ONNX format they can be added to your Nx AI manager.
- Finally, the model output is fed into the Nx metadata SDK, allowing you to use the entirety of the Nx system to collect data from multiple channels, visualize your data, and retrieve the associated video.
The Nx AI manager allows for much more complex “chaining” of models, channels, and output. However, in its basic form, an AI model that identifies an empty glass can easily provide a visual summary in the Nx client, showcasing all restaurant tables and pinpointing the customer in need of a refill the most urgently.
Scaling Your Solution
By leveraging the tools offered by Nx to build your Video + AI solution, you are primed to expand your solution across thousands of locations, devices, and channels in no time. While creating an initial PoC of a Video + AI solution is well within the skill set of an AI engineer, scaling and managing a video solution across a wide range of cameras and edge devices, each with their own specific AI accelerator(s), and sites is challenging for any skilled engineer. Drawing from a decade of expertise in developing highly scalable video solutions in various sectors, leveraging the Nx platform in conjunction with the Nx AI manager allows you to effortlessly deploy your solution on a large scale, in an instant.
How do I get started?
The Nx AI manager is set to be released alongside the first Gen 6 beta release later this year. If you are interested, there are two things you can do:
- Join our upcoming Hackathon. You will get first-glimpse access to the Nx AI manager and active support developing your solution by experienced Nx video + AI engineers.
- Reach out to our support team. If you're looking to build the next video + AI solution and conquer the world, we will get you started!
About the Network Optix 2024 Hackathon
The Network Optix 2024 Hackathon is a unique opportunity to develop enterprise video applications using the Network Optix Enterprise Video Platform and get an early look at the Nx AI Manager before its release. Participants will be challenged to create enterprise video applications that automate and enhance business operations through real-time video data analysis. With access to a range of video streams and cutting-edge development tools, it's not just a competition but a chance to contribute to transformative business solutions. It's a platform for developers to showcase their skills, collaborate with peers, and explore the potential of video technology in solving real-world problems.
Part 1 Begins: April 1, 2024 - Ends: May 1, 2024
Part 2 Begins: May 1, 2024 - Ends: May 25, 2024