Jeremy Roberson, Technical Director and Software Architect for AI and Machine Learning at Flex Logix, demonstrates the company’s latest edge AI and vision technologies and products at the 2023 Embedded Vision Summit. Specifically, Roberson demonstrates the power, performance, and area (PPA) capabilities of his company’s InferX IP by using it to run a state-of-the-art object detection neural network on four high resolution cameras in real-time.
In the realm of AI, deep learning models are constantly evolving. Yesterday, AI was dominated by CNNs, but today, generative AI and its transformer-based models are beginning to proliferate, especially at the edge. Any hardware designed to do efficient inference needs to be future-proofed enough to adapt to the rapidly changing landscape, while also maximizing its PPA. Flex Logix’s InferX IP is a unique architecture that combines hardened ASIC TPUs with a small amount of embedded FPGA fabric (EFLX) in order to maximize PPA while adapting for the future.
