Abstract

Autonomous vehicles rely on advanced perception systems to navigate and interact with their environment. This study presents a novel detection-segmentation network implemented on an energy-efficient SoC FPGA platform, achieving significant advancements in real-time obstacle recognition.

Highlights:

• Implementation on the AMD Xilinx Kria KV260 Vision AI platform
• High accuracy in object detection and image segmentation
• Significant reduction in power consumption compared to CPU-based systems

Future Directions:

• Investigation of scalability to different vehicle sizes and environments
• Enhancement of the platform’s adaptability to varying operational conditions

The achievements in this domain not only push the boundaries of autonomous vehicle technology but also create a roadmap for future developments in vehicle perception systems.