Solution
Smart eXtensible Vision Processing Unit
The intelligent solution for embedded vision applications in industrial vehicle technology.
From lab to field – proven by AMD
SXVPU is an AI-powered imaging platform designed to tackle challenges like object recognition and machine learning integration through its efficient real-time processing of high-resolution video streams up to 4K.
With its specialized AMD-based hardware architecture and support for multiple cameras and sensors, SXVPU is set up to use impeccable image data straight from the sensor for processing by neural networks or conventional algorithms before encoding the data for transmission via Ethernet.
The result is a perfect solution for viewing and vision applications in mobile automation.
You are currently viewing a placeholder content from YouTube. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More Information
Application Examples:
- Automation
- Monitoring of construction sites and work environments
- Collision avoidance systems
- Machine control
A Complete Solution for AI Vision Applications
Features:
- Flexibility, adaptability, robustness
- 24/7 safeguarded operation
- Low latency
- AI-powered for seamless integration of neural networks
- H.264/H.265 codec up to 4K video streaming via Ethernet
- High performance and efficiency in processing visual data
- Housing with IP67 protection
AI Development Support:
- AI model design
- Application development
- System implementation
SXVPU – Modular Approach for Ample Flexibility
System Performance
- High computing power and efficiency of up to 10 TOPS per watt
- H.264 / H.265 codec for up to 4K video streaming
- Low latency for real-time processing of video streams
- 24/7 safeguarded operation
Customizable HW
- Flexibility on all interfaces for various environments
- Scalable processing & versatile hardware acceleration
- Easy configuration and simplified system integration
- Housing with IP67 protection
Expandable SW + Intelligent Apps
- Seamless integration of neural networks and ML models
- Fast and efficient development with Linux® OS
- Tailored processing for industrial vehicle applications
- Support for common frameworks and industry standard
Optional Bird’s-Eye View Application
- Gives an overview of a vehicle’s surroundings
- Object detection algorithms
- Blending of image data from multiple sources
- Supports up to 8 cameras
Cyber Security Features
- Encryption and authentication of configuration files
- Hardened crypto-accelerators available for user applications
- Secure methods for storing cryptographic keys via eFUSEs
- Methods for detecting and responding to device tamper events
- OTA update authentication
Supported Standards
- TPM 2.0
- SHA-3/384
- AES-GCM-256
- RSA
Provisioning & Storage
The firmware is encrypted with a symmetric key that is itself wrapped using a device-unique PUF, so only PUF-bound, encrypted artifacts are stored in non-volatile storage. They are unusable on any other device.
Boot & Trust
During boot, the PUF reconstructs the key to decrypt and execute the firmware, making the PUF the root of trust and ensuring that plaintext keys and code exist only transiently at runtime.
Available SXVPU Models
AI Model
An AI-accelerated model that offers seamless integration of neural networks for AI vision applications. More information
Wi-Fi Model
A wireless system variant for use within a Wi-Fi network that meets the need for isolated operation. More information
For Quick Results: The SXVPU Dev Kit
Contents
- SXVPU
- 4x HDR camera
- IP69k protection
- 3 MPix at 60 fps
- Lens FOV H 118°
- 2 m cable
- Ethernet cable
- Power supply unit with cable
- Rugged carrying case
SXVPU Case Studies
With its flexible concept, SXVPU can be employed in a large variety of applications. Here are some practical examples for its use:
Project: Protected AI – Supervision of Construction Sites and Work Environments
Challenge:
The customer needs a safety-relevant and reliable person detection in demanding work environments. The solution should provide intrinsic safety regarding image acquisition and the potential neural network without substantially increasing the complexity.
Customer:
Manufacturer of construction machinery
Solution:
SXVPU enables the integration and safeguarding of LVDS cameras as well as a robust, independent supervision functionality outside of Linux, on top of the AI for high reliability and device safety. The system offers an easily implemented protection layer and simplifies error management.
Project: Smart Robotics Front-End for Automation
Challenge:
The customer has additional, crosscutting requirements for a camera system regarding hard real-time and flexibility of image acquisition. This goes beyond the integration of special camera heads, also concerning the precise control of exposure, shutter and further imaging parameters in the front-end in order to realize a comprehensive solution
Customer:
Mobile Automation – Robotics
Solution:
SXVPU enables the integration and safeguarding of LVDS cameras with individual additional functions and controls in the front-end. This flexibility ensures reliable and robust image acquisition, enabling not only the seamless integration of additional periphery on top of the camera heads, but also the direct realization of complex image analysis, substantially increasing the overall performance of the system.
Project: Lowest Latency Imaging for Real-Time Processing
Challenge:
The customer needs an intelligent camera system with minimum latency regarding video compression and direct connection to an AI, which is crucial for his overall application. Several use cases shall be supported, utilizing either a single high-resolution sensor or multiple HD sensors.
Customer:
Industrial vehicle technology
Solution:
SXVPU can be used to realize a low-latency connection of up to 4 cameras with a total resolution of 8 MPix to a neural network implemented within the unit itself. The latency-optimized transmission of compressed images via Ethernet is realized within 10 ms after image acquisition. This way, quick and precise processing of the image data is ensured, which is critical for the customer’s requirements.
Project: Agricultural Automation – Improving Efficiency through Intelligent Control
Challenge:
The customer plans to expand his portfolio with intelligent controls to improve the efficiency of agricultural processes or to realize their partial automation. A central task consists of implementing controls for precise trajectory control, loading management, and row detection in a self-sufficient box.
Customer:
Agricultural technology
Solution:
SXVPU offers a custom-tailored platform for multiple cameras and sensors paired with AI algorithms. The system enables the require precision and efficiency through its high flexibility in the choice of sensors and the seamless integration of neural networks in a self-sufficient box.
Project: IP65-based IPC for Machine Vision
Challenge:
The customer is looking for a high-performance AI-based computer vision solution with IP65 protection that meets the requirements for various environments, that is robust and can be expanded in flexible ways.
Customer:
Logistics automation – AGV/AMR
Solution:
SXVPU offers the optimal solution with support for up to 4 cameras, GbE LAN connection, expandability by EtherCAT, Profibus and real-time options. This way, the customer can use this IPC in different environments like smart city/industry, traffic management, logistics automation, and industrial vehicles.
SXVPU is a registered trademark of Solectrix GmbH. All other trademarks are the property of their respective owners.
