ADLINK and Intel Empowering Edge AI Developers from Prototype to Production with Vizi-AI™ and the Intel® Distribution of OpenVINO™ Toolkit

AI-powered machine vision technology is driving industrial transformation in manufacturing, employee safety and warehousing and logistics. Systems that can monitor and analyze live visual data can not only help in saving time, money and enhancing quality, but can even save lives. For example, in industries that deal with chemical or metal production, the use of AI-powered machine vision can help employees avoid injuries, provide warnings in case of an accident and design paths and processes for optimized and safe worker performance. In other cases, AI vision systems can identify whether workers are wearing Personal Protective Equipment (PPE) and report any non-compliance to safety protocols.

Enabling Outcome-Driven Solutions

The primary problem in designing machine vision systems for industrial purposes is the specialized nature of their design. The first task is to clearly define the outcomes which help developers zero in on determining the machine vision algorithms that will drive the solution. For instance, developing a solution that improves the quality of a finished product without impeding existing productivity levels.

Working backwards is a common approach to determining the requirements for an application. First, identify the problem that is preventing that outcome (e.g. too many weld errors leading to poor quality or unusable product.) Next, research technologies that provide solutions to that problem. And finally, start experimenting.

But how does a developer know what is possible with AI-based machine vision running at the edge? Edge AI provides many benefits, including lower costs from reduced storage requirements, increased security by keeping data on-premises, and access to real-time, actionable data through low latency needs.  But it is difficult to find one individual or even a team who has all the skillsets required to create a machine vision solution from scratch, and there are many other challenges to Edge AI, including networking, managing remote devices and camera and application integration.  

To solve this problem and enable developers to design, deploy and improve AI-powered machine vision solutions swiftly, ADLINK has created the Vizi-AI™ devkit supported by the Intel® Distribution of OpenVINO™ toolkit. Vizi-AI™ combines plug-and play industrial-ready hardware and software for a faster, easier and scalable starting point for machine vision AI deployments at the edge. Integration of the Intel® Distribution of OpenVINO™ enables optimization of deep learning workloads across ADLINK’s solutions. These solutions are powered by Intel® Movidius™ processors with performance that scales from the edge to the cloud and from prototyping to production environments. Vizi-AI™ is a simple way to get started to quickly help solve problems solution builders may face.

The Vizi-AI™ starter devkit includes an Intel® Atom® based SMARC computer module with Intel® Movidius™ Myriad™ X VPU. The addition of a 40 pin connector makes it compliant with Raspberry Pi HATS and compatible with Eclipse MRAA along with the Intel® Distribution of OpenVINO™ and ADLINK Edge™ software. From prototyping with ADLINK Vizi-AI™ to productionizing with ADLINK Neon smart cameras, or other industrial vision systems, this write-once, deploy-anywhere approach offers developers flexibility, scalability and streamlined integration.

Hit the Ground Running Right Out-Of-The-Box

Our developer platform is based on a configuration-over-code approach; we have built a UI that makes it easy to navigate and configure your Vizi-AI™ device with software-defined profiles, making it a no-code environment that is driven by configurations and best practices for the Intel® Distribution of OpenVINO™. 

Once developers choose a machine vision algorithm, they need to define or find a model that will help them achieve their desired outcome. This is where the Intel® Distribution of OpenVINO™ toolkit proves to be a valuable resource. Developers have access to a wide variety of pre-trained Intel models hosted on the Open Model Zoo. They can choose a model that fits their requirement or even train their model themselves using the ability of Vizi-AI™ to record images

Vizi-AI™ through ADLINK Edge™ also supports remote management and update for both the configurations and the model that has been established so that you can deploy, manage and update your solution when it is left on site for prototyping. Prototyping outcomes have recently been made easier with the introduction of the ADLINK Edge™ app connector for Node-RED, a programming tool for wiring together hardware devices, APIs and online services in new and interesting ways.

Seamless Transition from Prototyping to Productionizing

ADLINK and Intel technologies help make the transition from prototyping to productionizing seamless. The software platform and communication that takes place between the applications play a major role in this transition and is enabled by the ADLINK Data River™, a data sharing platform that enables the ability to plug-in modular software and hardware anywhere at the edge to facilitate North-South and East-West data flow. This modular self-discovering capability allows applications to be deployed onto other hardware with more suitable processing or storage capabilities for the solution as it evolves without changing the application software.

The Intel® Distribution of OpenVINO™ helps developers bypass a key deployment issue. When designing a solution, developers need to keep in mind the computational resources they have available for the execution of the model. Going from the hardware on which the model is developed to the actual embedded hardware for deployment is often a big leap and a point of ambiguity for developers. Our Vizi-AI™ solution eases the transition by making the target platform and the development platform appear similar and by dealing with the abstractions. Models built with the Intel® Distribution of OpenVINO™ are equipped to make the best use of the hardware that they run on. Even if they are deployed on hardware with low computation capabilities, these models can recognize when they have additional computational power available and enhance their performance accordingly. 

When it comes to the performance of the model and understanding it, the Intel® Distribution of OpenVINO™ toolkit comes with a Deep Learning Workbench (DL Workbench), which allows developers to examine the performance of their model on different hardware. Once the deployment hardware has been selected, developers do not need to change their software when they transition from Vizi-AI™ to a larger production and performance system.

The Edge and Other Advantages

ADLINK Edge™ software is an accelerator to get you up and running much quicker. If you are a developer and you don’t want to build everything yourself, ADLINK Edge™ has a number of off-the-shelf applications, including connections to different kind of cameras, that will make it easier for you to get started with Vizi-AI™.

The Vizi-AI™ platform works closely with the ADLINK Data River™, offering highly deterministic, forgiving and scalable low-latency data distribution with automatic discovery and built-in security. This technology is high-performing with latencies as low as 30 µs and a throughput of tens of thousands of messages per second, proving to be well suited for almost every use case. It is a proven technology with a background in military and aerospace based on the open Data Distribution Service (DDS) standard.

The Vizi-AI™ solution also comes with access to a set of data models that are used to normalize data coming from different machine vision models. No matter what model is run for classification, the output data will be consistent. This saves the developer the trouble of having to understand the underlying architecture and the way it represents the predictions and inferences produced by the solution. Not all classification algorithms produce results in the same manner. The same logic applies to all object detection, segmentation and other machine learning models: they produce results differently. The advantage of using Vizi-AI™ is the set of data models that allow us to normalize this data, making it easier to regulate the applications constructed against those data models.

Making a Difference with Machine Vision Technology

Our simple, scalable devkit makes it easier for developers to take a machine vision solution from the prototyping stage to the factory floor to test out the original desired outcome. With the ADLINK Edge™ Software and Intel® Distribution of OpenVINO™ toolkit, the solution is quickly ready for deployment on correctly sized and industrialized hardware. This devkit empowers developers to deploy solutions faster and provides businesses with their desired outcome, such as the ability to improve product quality, improve worker safety and reduce downtime. 

Where to Get Yours:

Vizi-AI™ Industrial Machine Vision AI Developer Kit— Build your solution with this easy-to-use kit with plug-and-play hardware and software that provides a faster, easier and scalable launchpad for machine vision AI deployments at the edge.

ADLINK Edge SDK™— If ADLINK doesn’t have the app, then unlock the data in the Data River with the SDK.

Intel® Distribution of OpenVINO™ toolkit— Explore how you can take advantage of the write-once, deploy-anywhere approach.

Join the Community

Vizi-AI not only brings you collaborative solutions but also a forum for collaboration and learning where you can interact with fellow developers and access information and resources to help you through your machine vision development cycle. Join the community today. 

_{Notices and Disclaimers}

_{Intel technologies may require enabled hardware, software or service activation.
No product or component can be absolutely secure. 
Your costs and results may vary. 
© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.}