The SXVPU Dev Kit: Real-World Edge AI Use Cases

Thanks to its flexible architecture, the SXVPU is suitable for a wide range of applications.

The SXVPU is an AI-powered imaging platform designed for demanding machine vision applications, including object recognition and machine learning. It delivers efficient, real-time processing of high-resolution video streams up to 4K, enabling fast and accurate AI inference at the edge. Built on a specialised AMD-based hardware architecture, the SXVPU supports multiple cameras and sensors, processing high-quality image data directly from the sensor using neural networks or conventional vision algorithms before encoding and transmitting the data via Ethernet. The result is a high-performance, low-latency solution for intelligent vision and imaging applications in mobile automation, industrial systems, and other edge AI environments.
SXVPU Dev Kit Store Page

Here are some real-world examples of how the SXVPU can be used:

Protected AI – Construction Site & Workplace Supervision

Customer

Construction machinery manufacturer

Challenge

The customer required a reliable, safety-critical person detection solution for demanding construction site environments. The system needed to provide dependable image acquisition and AI-based detection while ensuring functional safety without significantly increasing system complexity.

Solution

The SXVPU enables the secure integration and monitoring of LVDS cameras while providing an independent supervision layer that operates separately from Linux. By complementing the AI-based detection system with robust safety monitoring, the solution enhances reliability, simplifies error management, and provides an additional layer of protection for safety-critical applications.

Smart Robotics Front-End for Automation

Customer

Mobile automation and robotics manufacturer

Challenge

The customer required a camera front-end capable of meeting demanding hard real-time performance while providing the flexibility to integrate specialised camera heads. In addition, the solution needed precise control over exposure, shutter timing, and other imaging parameters to support advanced image acquisition and enable a comprehensive, high-performance vision system.

Solution

VPU enables the seamless integration and safeguarding of LVDS cameras while providing flexible front-end control of imaging parameters and additional camera functions. This allows specialised camera heads and peripheral devices to be integrated with ease, ensuring reliable, deterministic image acquisition. By supporting complex image analysis directly at the front-end, SXVPU enhances system performance, increases flexibility, and delivers a robust solution for advanced mobile automation and robotics applications.

Lowest-Latency Imaging for Real-Time Processing

Customer

Industrial vehicle technology manufacturer

Challenge

The customer required an intelligent camera system capable of delivering ultra-low latency between image capture, video compression, and AI processing. The solution needed to support multiple use cases, accommodating either a single high-resolution image sensor or multiple HD cameras while maintaining real-time performance.

Solution

The SXVPU provides a low-latency interface for up to four cameras with a combined resolution of up to 8 MP. A neural network running directly on the device enables immediate AI processing, while latency-optimised Ethernet transmission delivers compressed image data in as little as 10 ms after image acquisition. This ensures fast, accurate image processing for time-critical industrial applications where real-time performance is essential.

For more details, email us at sales@sundance.com or view the SXVPU Dev Kit Store Page below:
SXVPU Dev Kit Store Page