Launched in 2021, VisualCortex develops a Video Analytics platform that runs computer vision models developed by VisualCortex or third parties. The platform helps enterprises improve the security, safety and efficiency of their operations. Based in Sydney, Australia, the company works with clients in many industry sectors, from retail and hospitality to transportation, manufacturing and government.
In the past, video analytics solutions have often been prohibitively expensive and complex, with relatively low levels of accuracy. And while many developers have launched specialized point solutions, few options were available to tackle multiple use cases.
VisualCortex aims to change this by delivering powerful video analytics with over 95% accuracy that support a wide range of use cases, from theft prevention and inventory management to promotions planning and workforce optimization.
At-a-Glance
VisualCortex uses SUSE Rancher Prime to simplify the orchestration, deployment and remote management of its containerized, AI-driven machine learning (ML) models, supporting its mission to make sophisticated video analytics and intelligence more accessible. When a major retailer requested new car park management functionality, VisualCortex added SUSE Storage to record data on thousands of incoming and exiting vehicles each day.
The journey to SUSE and containerization
VisualCortex’s core platform enables users to orchestrate, monitor and deploy its AI-powered computer vision models across multiple locations. Once connected to live or archived video footage, the models identify events and behavioral anomalies, before alerting stakeholders to take prompt and effective action — whether that’s to prevent crime, optimize store layouts and manufacturing processes or enforce health and safety protocols. The platform uses AI frameworks such as ONNX, TensorFlow and NVIDIA inference technologies.
When they began developing this platform, VisualCortex recognized that containerization was essential for the scalability, efficiency and reliability needed to support AI/ML workloads. To enable this strategy, the company opted for Kubernetes, implementing RKE2 and K3s distributions to run the containerized workloads.
Using an edge computing architecture means that VisualCortex often has very little processing power for deployment. Some edge devices are powered by batteries or small solar panels. To optimize energy consumption, VisualCortex needed the smallest possible software footprint. The lightweight nature of K3s allows it to be easily deployed on any architecture, maximizing performance while minimizing energy consumption.
In turn, using containers created additional administration challenges for the VisualCortex team. Ben Evans, Chief Technology Officer at VisualCortex, explains: “Kubernetes can be a complicated beast. Because we use the edge computing architecture when deploying to clients, we need to coordinate containerized clusters across many different physical locations, which adds extra layers of complexity.”
“Rancher Prime provides the flexibility and scalability we need to run AI/ML workloads seamlessly across edge and cloud environments.”
Why SUSE solutions?
Why Rancher Prime?
To support the delivery of its AI-powered video analytics platform, VisualCortex looked for a powerful system to simplify the management and monitoring of its central Kubernetes deployment and edge clusters. After reviewing various container orchestration solutions, VisualCortex initially selected open source Rancher.
However, although the VisualCortex team found that Rancher Community Edition met all of its requirements, some of the company’s enterprise clients wanted the additional reassurance of enterprise-grade support for their complex production environments. VisualCortex therefore upgraded to Rancher Prime for these clients.
“Rancher Prime provides the flexibility and scalability we need to run AI/ML workloads seamlessly across edge and cloud environments,” adds Evans. “Many container management solutions are tied to a single cloud provider with high costs. We needed an AI-optimized, cloud-agnostic platform that could support AI/ML workloads at scale across multi-cloud, edge and even air-gapped environments. Rancher Prime hit the mark perfectly.”
In addition to its robust management capabilities, Rancher Prime includes built-in security features such as centralized authentication and access control, helping VisualCortex meet stringent security requirements.
Today, Rancher and Rancher Prime serve as the main container orchestration layer in the VisualCortex platform. The SUSE technology enables the company to manage the central Kubernetes control plane and edge components, including Google Kubernetes Engine (GKE) — quickly and without hassle or complexity.
“Rancher Prime is reliable, extensible and enables simple administration of containerized clusters deployed across multiple physical locations,” adds Evans. “Any time that we win new clients, or our existing users add new sites, Rancher Prime makes it easy for us to deploy our computer vision models to Kubernetes clusters running at the edge.”
Rancher Prime has also been vital in delivering high-quality support to clients. As Evans explains: “We can handle some troubleshooting requests for the third-party technologies in our stack, but many of our larger clients require enterprise-grade support with stringent SLAs. Rancher Prime means that we can deliver just that.”
Why SUSE Storage?
Recently, VisualCortex received a request from a major Australasian retail group to create a vehicle recognition engine (VRE), based on its core platform. Part of the VRE project was a car park management system that would use computer vision to automatically identify and log incoming and exiting vehicles. The main objectives were to replace error-prone human monitoring, and to improve the speed and reliability of parking revenue collection.
The development and launch of the VRE introduced a new challenge for VisualCortex: the solution required highly scalable, resilient and low-latency storage capabilities to efficiently record details of the thousands of vehicles visiting the retailer’s locations daily.
Building on its positive experience using Rancher Prime, VisualCortex selected SUSE Storage to provide storage resources for the car park management system. The company also engaged SUSE Consulting to help validate the architecture and design of the new solutions, and to support the implementation process at several of the retail group’s sites.
Within the car park management system, SUSE Storage plays a critical role, storing a database that records data on all vehicles that enter and exit each retail location, including their arrival and departure times and registration plates. Using this information, the retailer can then automatically calculate parking fees for collection.
Evans comments: “In our experience, persistent storage is one of the toughest things to get right with Kubernetes. In the past, we have used cloud-based block storage, but that wouldn’t make the grade this time, as we were storing data at the network edge. SUSE Storage was the ideal fit, a secure solution that delivers high availability, scalability and redundancy.”
The impact of Rancher Prime
Supports a growing client base
SUSE Rancher Prime has played a key role in enabling VisualCortex to deliver its core platform to enterprises across many industry sectors. Once clients sign up, the company can deploy the platform at their locations quickly, enabling immediate analysis of video sources and providing detailed insights to on-site stakeholders.
“Rancher Prime is an amazing tool, making what would normally be challenging container management tasks simple,” adds Evans. “We can easily link our central control plane to edge clusters, irrespective of where they are physically located. Once up and running, it’s easy to upgrade both VisualCortex and Kubernetes components.”
At the same time, the streamlined management of clusters does not restrict VisualCortex’s choice of Kubernetes provider, as SUSE Rancher Prime integrates seamlessly with the company’s K3s, RKE2 and Google Kubernetes Engine clusters.
“K3s works on every architecture,” says Evans. “Choosing Rancher Prime has really given us the freedom of choice we need, whether we deploy K3s or RKE2. It’s so valuable, especially for the power it gives us to quickly iterate solutions running on remote systems.”
Underpins effective support for enterprise clients
In addition, Rancher Prime enables VisualCortex to address technical issues quickly and effectively. “The web-based interface and granular permissions structure in Rancher Prime allows our team to isolate and troubleshoot problems for clients, without having to access any other sensitive workloads running on the same infrastructure,” adds Evans.
The impact of SUSE Storage
Enables launch of innovative new capabilities
Using SUSE Storage, VisualCortex successfully developed and launched the VRE system, including the crucial car parking management component. Currently, the company has rolled out the VRE solution at four of the major retail group’s locations, where the system collects and analyzes video feed data from hundreds of cameras.
Evans notes: “Much like Rancher Prime, SUSE Storage has been extremely reliable, enabling us to focus on what’s important: building new computer vision models rather than low-level administration. It’s easy to provision and manage, and the automated backup function is very useful, too, saving our team valuable time on routine work.”
Achieving exceptionally accurate video analytics
The new VRE functionality has already yielded highly positive results for the major retail group.
“We have achieved extremely high accuracy with the VRE and car park management system,” explains Evans. “That has enabled the retailer to reduce the number of manual corrections required to read vehicle plates and process parking charges, contributing to an increase in revenues collected. Without Rancher Prime and SUSE Storage, we simply wouldn’t have been able to move the VRE system into production.”
The high accuracy of the VisualCortex platform also supports the retailer with reducing customer support complaints. With human error minimized, people are less likely to get mischarged for their parking, and the customer support center has already seen a significant reduction in calls.
What’s next for VisualCortex?
Innovation is part of VisualCortex’s DNA, and the company is continuously developing new AI/ML models and video analytics solutions. For example, the team is currently researching and trialing new AI-based methods for enhancing public safety and detecting violence, as well as for commercial use cases such as retail store journey mapping. One retail client plans to analyze vehicle license plates to determine the proportion of electric vehicles, using that data to plan charging stations in car parks.
Evans concludes: “In some cases, we lead the development work, and in others we work with a network of established partners to ensure we continue to enhance our solutions. As we expand our AI capabilities, SUSE will play a central role in our ongoing operations. We have had an excellent experience using Rancher Prime and SUSE Storage, and they have helped us to scale our portfolio and bring our groundbreaking video analytics platform to many more enterprise clients.”