Chetan Khona, Director of the Industrial, Vision, Healthcare and Science Markets at Xilinx, presents the “New Adaptive Module for Vision AI at the Edge” tutorial at the May 2021 Embedded Vision Summit.
Xilinx Kria System-on-Modules (SOMs) provide a secure, production-ready multi-core Arm and FPGA platform, including memories, power management, and Yocto or Ubuntu Linux to build accelerated AI-enabled applications at the edge. Kria SOMs enable smart vision applications across cities, factories, and hospitals to achieve high performance with low latency, low power consumption and a small footprint. And Kria SOMs feature a radical new way to design with Xilinx: Accelerated Applications via the Xilinx App Store.
Kria Accelerated Apps offer an industry-first: they enable both new and experienced Xilinx designers to skip doing any FPGA design. Accelerated Apps give the SOM the personality of a purpose-built smart camera, AI box, or other vision AI system. Apps are fully accelerated—including image acquisition, pre-processing, AI inference, post-processing, encoding and connectivity—offering the highest performance for industrial use cases. Accelerated Apps span license plate recognition, retail shopper re-identification, HDR image signal processing, natural language processing and more.
See here for a PDF of the slides.