Edge AI enables industry to make intelligent decisions close to the source of data. It is our passion at ADLINK and Vizi-AI shows how anyone can get started with Machine Vision AI solutions. How did it come about?
Let me take you back over the last four years. ADLINK acquired a software business that had built distributed data connectivity middleware for 25 years. This software issued in mission critical systems primarily in the defence and aerospace sector to allow data to move from one application to another across networks to build a system; e.g. Air Traffic Control Systems, Naval Combat Management Systems. The software team was well experienced in building software suitable for enterprise and embedded deployments supporting well over 45 different flavours of operating system and all the major compute architectures. We understood the level of quality and robustness required for distributed complex systems.
ADLINK’s history started in data acquisition modules and over its 25-year history it had become one of the world’s major industrial hardware manufacturers and IoT was becoming a growth area with the cloud vendors. Industrial IoT was creating high volumes of data that could create new high value revenue streams for the cloud.
Surely, by taking its hardware, now including IoT specific devices such as gateways and adding the data connectivity software backbone it had acquired, it would create extra value for its customers.
The next step was to build out its connectivity end points, specializing its mil/aero software for IoT and rebranding as the ADLINK Data River alongside the overall platform ADLINK Edge. A user experience was created to ensure the hardware/software could be deployed rapidly and easily updated remotely and ADLINK began implementing some real-world use cases, solving real problems concentrating on two areas.
Machine Health – connecting machines, traditionally in manufacturing, where customers are looking at Overall Equipment Effectiveness and combining ADLINK hardware / software to unlock machine data and move it to another application to create value either at a dashboard locally or the cloud for further analysis.
Machine Vision – ADLINK chose to focus first on Automated Object Inspection within manufacturing and the opportunity to show how the power of AI could allow the same hardware/software infrastructure to be applied in many more use cases, simply through training rather than programming.
Using the experience of the software team, we built these solutions by creating small modules of software that when put together would create an end solution. This maximised re-use and due to the power of the Data River, meant they could be on one device or many without any changes to the software modules or any configuration. This allowed the computing resources to be managed much more effectively at run time and enabled systems to evolve very easily.
Which brings us back to why we created Vizi-AI.
We built the machine vision solutions in a very generic way. The part that changed was either the camera required (due to the environment, lens, etc) or the AI component to make its decision on what the camera was seeing or the hardware architecture to get the right performance. This meant we could build different systems by using our vast hardware range and combining the software components together using our catalogue of camera technology apps or changing the vision AI model. No programming, just configuration and AI model training.
With Vizi-AI we have created a starting point. We’ve provided a development hardware/software compute environment that allows new users to get familiar with the software and to easily prototype solutions, knowing once they are ready to deploy, then utilising the same software infrastructure they can make hardware choices at deployment time. This is what we have done in ADLINK in the real world, using our experienced hardware and software team to ensure it is real-world-ready and giving new users the benefits of our experience. It isn’t the silver bullet, as there is always the last step to complete with Vizi-AI; what do you do with the result of the AI, but we’ve provided the ability for users to do that part of their journey in their solution, by making access to the data freely available.
Vizi-AI is about showing the art of the possible and with its supporting goto50.ai community to get assistance and ideas for innovation.
What challenges does this address, well as the leader of a business that itself uses 3rd party components in its solution, I often face the following challenges;
- Support: Security of supply and the insurance that when there is a problem I know where to go to get the support required. At ADLINK, as its both the hardware and software that is supported, there is 1 point of contact for a lot of the solution stack.
- Build vs Buy: Do I try and build this myself or do I just buy it. I could plug together open source software and add some glue code, but is that what my organisation wants to specialise in, is that creating value to my customers? From experience, knowing that the underlying infrastructure I have chosen is supported and validated and most importantly kept up to date and market relevant. It is often under-appreciated how much time is taken up in configuration management and validation to ensure that the infrastructure is ready and in IoT, with security key, this can become a significant burden. By buying, you share that cost with all other users of the infrastructure and also gain from a steady stream of product updates.
- Time to market: Related to build vs buy, but given how fast new disruptive solutions are being offered, time to market has to be considered.
- Skills and knowledge risk : Do I have the people to put the complete solution myself and the coverage for knowledge loss or do I just create the commercial partnership to allow me to focus on the aims of my business not the standardised components. Do I even have the experts? there is a major shortage of skills and retention challenges.
- Technology choice: Does my choice of commercial partner have the right relationships to ensure my product will be competitive and flexible in the market. With ADLINK’s software stack supporting multiple hardware architectures and ADLINK having strategic relationships companies such as Intel, Nvidia, then your choice for your customer installation is more flexible and brand acceptable. Additionally, when new product arrives, ADLINK will be bringing it to market to give you competitive advantage.
- Open: Am I being locked into anything that will penalise me later. The ADLINK Data River technology is open source, ensuring that at all times, access to the data in the system is available for use, with an easy to use API designed to allow users to extend their system without being dependent on ADLINK.
- The importance of non-functional requirements – What about scalability, reliability, fault-tolerance, determinism, extensibility, ease of integration, etc? Why starting off with the right technology matters when building real systems, opposed to building a PoC.