Technical Presentation: In-Sensor AI Computing for Efficient Real-Time Machine Perception
Artificial intelligence (AI) and machine learning (ML) technologies have fueled many burgeoning applications from on-device learning to self-driving cars and collaborative robots. Despite these unprecedented advancements, the holy grail of enabling fully autonomous machine intelligence remains far from our grasp. One key challenge is the lack of performant and efficient hardware implementation, especially in the case of embedded/edge devices with rich sensory inputs yet stringent resource constraints.
In this talk, I will present our work to tackle this challenge from a unique angle that leverages information processing ability innate in the analog/mixed-signal (AMS) domain by embedding AI computation directly inside the pixel circuits, delivering much-improved performance and efficiency. I will first introduce our design of in-sensor learned compressive acquisition (LeCA) and then talk about a novel in-sensor architecture that supports real-time gaze tracking, a key primitive in AR/VR applications. Finally, I will present CamJ, a first-of-its-kind energy modeling framework that empowers designers to navigate the large algorithm-hardware co-design space of computational image sensors.
Speaker(s): Silvia,
Agenda:
Technical Presentation: In-Sensor AI Computing for Efficient Real-Time Machine Perception
Virtual: https://events.vtools.ieee.org/m/431983