Scalable, secure and swift. NVIDIA’s range of boards has proved revolutionary in the field of machine learning and artificial intelligence.
With powerful hardware capable of performing on the edge of AI systems and software that adapts to different challenges at pace, it’s little wonder why NVIDIA has become a go-to choice for so many looking to start projects that involve autonomous processes.
In our guide, we’ll take a look at exactly what makes NVIDIA so capable of underpinning these systems and investigate some case studies to see it all in action.
If you think NVIDIA’s platform could provide the foundation for an AI solution for your business, check out our selection of boards and accessories or speak to one of our experts for more information.
What makes NVIDIA great for AI?
- Processing power: NVIDIA’s Jetson range delivers 40-275 TOPS (trillion operations per second) across a series of boards that can comfortably process information from multiple sensors before turning it into real-world data for either you or the processor to analyse. All that power doesn’t come at the expense of space, however. The powerful Jetson AGX Orin™ 64GB Developer Kit measures just 110mm x 110mm x 71.65mm but delivers industry-leading performance.
- Development tools: As well as heavy-hitting hardware, the NVIDIA Jetson software stack is compatible right across the range of boards to make projects more adaptable and agile. Data can be transferred across projects to prevent the need to create every model from scratch.
- Edge capabilities: Computation performed closer to the ‘edge’ of the network, i.e. closer to where devices and sensors actually are, has proved game-changing for AI capabilities. Quicker data capture and response times, less cloud-storage costs and improved security open more doors for AI to improve more processes in our daily lives.
NVIDIA Case studies
OK, so we’ve told you how great NVIDIA is for AI applications. If that doesn’t impress you much, why don’t we show you instead?
Edge AI improves outcomes for dialysis patients
The problem
Dialysis patients are at great risk of heart complications due to the links between the cardiovascular system and our kidneys. Consultants aim to return patients to their ‘dry weight‘ – how much they weigh with excess fluid removed from the body – however, this process can put a strain on the heart and result in high blood pressure.
The solution
Taiwan’s Taipei Veterans General Hospital (TVGH) sought to mitigate this with an AI model that could predict heart failure while patients were undergoing dialysis. Understanding which patients were at greater risk would allow TVGH to deliver specialised care to those who needed it most, with doctors and nurses made aware when a patient was displaying abnormal patterns in their heart.
Photo credit: nvidia.com
The kit
- NVIDIA Jetson Xavier NX
- NVIDIA CUDA-X libraries
- NVIDIA TensorRT software
The outcome
Trained on a dataset of three million medical records and bolstered by a neural network model, the AI tool reached 95% accuracy when reading patients’ vital information and predicting their future condition. An initial need for human input led to a 30-minute delay in reporting – a potentially vital window for medical staff – so NVIDIA’s edge capabilities were put to the test, bringing this wait into the milliseconds as a result.
You can see this innovative healthcare solution in this video here.
Robotic farmhands create a new tech sector?
The problem
Farming is labour-intensive and yields are often unpredictable. This is especially true of root vegetables that grow underground such as carrots. Other produce like onions, apples and broccoli also present issues with short windows where fertilisation and harvest are viable for them to be sold to consumers.
The solution
A San Francisco-based startup, Verdant, aimed to create an AI-powered 3D visualisation system that could provide the bedrock for tasks such as weeding, fertilising and spraying to be performed by robotic tools. Are you already familiar with SaaS? Welcome to RaaS – Robots as a Service.
Photo credit: verdantrobotics
The kit
- NVIDIA Jetson AGX Orin
- TAO Toolkit
- DeepStream
The outcome
Working at a Bakersfield, CA carrot farm, the machine’s weeding capabilities greatly reduced the need for herbicides to be used, having an immediate positive effect on the farm’s sustainable and regenerative goals. Improvements aided by the NVIDIA TAO toolkit also helped Verdant adapt existing models five times quicker – for example, tweaking carrot detection so it can work for onions in other settings.
You can see the case study video from Verdant Robotics here.
“We display a model of the tractor and a 3D view of every single carrot and every single weed and the actions we are doing, so it helps customers see what the robot’s seeing and doing,” says Verdant co-founder Lawrence Ibarria.
Warehouse AI technology
The problem
PepsiCo’s distribution warehouse requires packages to be manually picked and labels scanned as orders are put together. Fixed to cardboard boxes, these labels could be upside down, obscured from view or even damaged and difficult to decipher. This leads to increased time spent in the warehouse finding the component parts of an order for distribution.
The solution
NVIDIA Metropolis partner KoiReader Technologies developed a barcode reader designed to validate customer orders, ensuring that no products were missing and quantities were correct. The reader was able to decipher the barcodes, no matter what angle they were at and even if some of the label was obscured or torn away.
Photo credit: nvidia.com
The kit
- NVIDIA Jetson AGX Orin
- NVIDIA TAO Toolkit
- NVIDIA DeepStream
- DALI
- TensorRT
The outcome
A simply flawless 100% accuracy of human-assisted picking processes, with the capability to scan barcodes as they whizz past the scanner on conveyor belts. Originally tested at PepsiCo’s Dallas-Fort Worth facility, the technology is set to be introduced in more locations on the back of its success.
Develop groundbreaking AI solutions with NVIDIA and OKDO
Have you got a great idea for the future of AI that you want to bring to life? NVIDIA’s high-power development boards and software could unlock its potential. Discover the latest trends in industrial AI and the possibilities that OKdo and NVIDIA bring to your business in our recent webinar.
We stock a wide range of NVIDIA boards and accessories to help you get started. Our Projects Hub and Blog are also packed full of ideas that might serve as inspiration for your next venture.
OKdo Nano C100 Developer Kit powered by NVIDIA® Jetson Nano Module
NVIDIA® Jetson AGX Orin™ 64GB Developer Kit
NVIDIA® Jetson Orin™ Nano Developer Kit
Need extra information on how we can help? Feel free to contact us and our experts will be in touch to talk through the particulars of your project.
