Visidon uses LUMI supercomputer to handle peak workloads
Visidon is an excellent example of how smaller technology companies can also benefit from the LUMI supercomputer. Although LUMI is one of the most powerful supercomputers in the world, its use is not limited to large corporations with massive datasets. It offers a flexible and cost-effective computing environment for smaller companies that need speed and scalability for product development.
Founded in 2006 as a spin-off from the University of Oulu, Visidon currently employs about 40 professionals specializing in image processing and machine learning. Its main markets today are Asia and the United States.
Visidon develops advanced, optimized software solutions for imaging and video quality enhancement, particularly in embedded electronic systems where computing power is often limited and real-time processing is required. Their technologies leverage AI and machine learning, focusing on visual improvements in image and video quality, better resolution, depth calculation, as well as facial and object recognition and tracking. Visidon’s solutions are well-suited for smartphones, robotics, video conferencing, and surveillance systems. The company’s technology can be found in over one billion mobile devices worldwide.
Visidon has its own servers for development work, but sometimes its computing capacity is insufficient. Servers are extremely expensive, and it doesn’t make sense to purchase them just for occasional capacity needs, leaving them idle most of the year.
Easy LUMI adoption – also without prior HPC experience
A few years ago, Olli Silvén, an emeritus professor familiar to Visidon from the University of Oulu’s Machine Vision and Signal Analysis Research Center, suggested that Visidon could use the LUMI supercomputer in Kajaani alongside its own servers. Inspired by Silvén’s tip, IT Systems Manager Mika Helistekangas contacted CSC’s Mikko Kerttula. Based on their discussion, a Try & Buy project was set up for Visidon, allowing them to test LUMI free of charge.
“The Try & Buy project includes user accounts for two users, CPU, GPU, and data storage resources for testing, and free expert support for onboarding,” explains Kerttula, CSC’s Senior Project Manager responsible for Visidon’s account.
“After the 3–6 month test project, the company can continue using LUMI by signing a computing services agreement. If LUMI doesn’t prove useful, they can simply stop after the test phase – at no cost.”
Visidon’s onboarding went smoothly.
“LUMI documentation is very clear, and we got started easily by following it,” says Helistekangas. “Also, using LUMI has turned out to be simple, even though we had no prior experience with high-performance computing.”
We use LUMI to finalize our products when our own servers are overloaded.
Mika Helistekangas, Visidon
After the Try & Buy project ended, Visidon continued using LUMI as a purchased service. They created their own documentation based on LUMI’s materials, making it easy for new employees to use.
“We use LUMI to finalize our products when our own servers are overloaded,” Helistekangas explains. “We do research and training mainly on our own servers because that’s iterative work. Then, when the research is done and the product needs to be finalized and completed, we switch to LUMI.”
LUMI enables parallel training on multiple GPUs, significantly speeding up processes. Training a single model can be 2–4 times faster than on the company’s own servers, and parallelization adds even more power.
Helistekangas praises LUMI’s Slurm workload manager, which allocates computing resources to tasks, queues, and priorities for efficient use. It simplifies management and accelerates job execution on supercomputers.
According to Visidon, the greatest benefit of using LUMI is that it enables flexible customer service and helps them meet even the tightest deadlines.
Bright prospects ahead
How does Visidon’s future look like? Will the demand increase in Europe, even though exports currently go 100% to Asia and the United States? There’s much talk about Europe’s self-sufficiency.
“The more you think about it, cameras are everywhere nowadays,” reflects Jenna Enbuska, Visidon’s Marketing Director. “IoT and telecommunications technologies rely on cameras, as do drones and other autonomous vehicles, which are becoming more common and diverse across industries. The importance of image and video quality will only grow in the future. So we see the market outlook as bright and the need for LUMI continuing.”
For European companies, sustainable operations are important, so using the LUMI supercomputer in product development could be a selling point. Located in Kajaani, LUMI is one of the world’s most environmentally friendly supercomputers. It runs entirely on renewable energy, and the waste heat from its data center is used in Kajaani’s district heating network.
“The future looks even brighter knowing that in 2027, LUMI will get an AI-optimized successor, the LUMI-AI supercomputer,” Kerttula adds. “This means that companies’ access to high-performance computing for R&D is secured well into the future.”
Photo: Visidon
Mikko Kerttula
Mikko Kerttula promotes LUMI industrial use especially in Northern and Eastern Finland. In addition, he supports the development of the Kajaani data center ecosystem.
