Embedded Deployment for Real-Time Road Sign Detection and Identification

This article reports on the design and deployment of a Deep Neural Network (DNN) in a hardware platform for real-time detection and identification of road signs. A YOLOv3 DNN is trained using the Keras framework and deployed into an AMD Kria KV260 development board using the Vivado, Vitis IDE and Vitis AI tools. The DNN is implemented in a Deep Learning Processing Unit (DPU) core and the software application is developed at an operating system level using the Python Productivity for Zynq (PYNQ) framework from AMD.