SmartCow has announced it is integrating new Nvidia edge solutions to offer high-performance solutions for computer vision applications. The company has added two new edge products for video analytics and IoT use cases — AI SuperCam and the Mars box PC, both incorporating Nvidia Jetson AGX Orin system-on-module (SoM).
SmartCow develops customizable AI products to provide full-stack, multi-utility software to let enterprises connect, control, and coordinate large networks of edge-embedded hardware.
The SmartCow SuperCam is an IP65 rugged AI camera, designed to be deployed in harsh environments, such as transportation management. The camera module has capabilities to collect and process video edge data to reduce latency and increase the adoption of edge servers. The hardware operates in remote locations with limited internet connectivity.
On the inside, SuperCam is equipped with a 4K rolling shutter, fixed lens camera sensor and a global shutter sensor with a motorized lens. In terms of wireless connectivity, the hardware platform includes support for 10-gigabit ethernet and LTE/5G networks.
SmartCow Mars AI embedded box PC is built using the 32GB and 64GB variants of the Nvidia Jetson AGX Orin system-on-module. The hardware employs out-of-band power cycling and features remote debugging capabilities. This means the device can be managed even when the device does not have a network connection, which can help avoid (or recover from) a system crash.
Design challenges: data and security
From a computer vision point of view, SmartCow sees data as a challenge for the field of computer vision systems. Data forms the basis of the images that these systems are used to identify.
“Datasets are a challenge for building a computer vision project. Nowadays, privacy concerns are raised, and it gets harder to prepare a real-world dataset. Therefore, synthetic data generation will be a tech and AI development trend,” says Alice Lai, who heads public relations and marketing for SmartCow in an interview with EdgeIR. She quotes Gartner statistics that suggest that by 2024, 60% of the data used for the development of AI and analytics projects will be synthetically generated.
The challenges with data extend to security, as well, but edge devices in conjunction with 5G networks will help with security.
“At first, people chose edge computers instead of cloud services because they wanted to ensure data control. Furthermore, edge computing connected with 5G can also raise security. The 5G nowadays has a private connection, bringing a better security concern to the market,” Lai says.
As far as hardware and software design is concerned, SmartCow is banking on Nvidia for several reasons.
Lai notes that SmartCow has focused on edge AI end-to-end solutions since 2016. The company’s Software-Defined Hardware (SDH) provides field programmability and flexibility in hardware allowing the implementation of time-sensitive features to avoid potential system latency. The company has focused on Nvidia because “Jetson modules not only have unparallel computing capability but also feature their wide range of software developer kits (SDK),” she says. Nvidia SDKs, including Metropolis for vision AI for smart cities, Isaac for robotics, and Clara for medical, allow for quicker access to edge AI development. “People no longer struggle to start in an early stage,” she notes.
As to the hardware, “Jetson is a leading module with outstanding computing performance and different levels of modules. It lets us develop different applications to meet the market’s diverse needs,” she says. Also, “Jetson is fully integrated with a wide range of SDK, and our engineers can work on multiple AI models seamlessly,” she adds.
5G | computer vision | device management | martCow | Nvidia | video camera