Towards Ultra-Low Power Embedded Object Detection
Explore ultra-low power embedded object detection in this tinyML Talks webcast featuring Theo Theocharides from the University of Cyprus. Delve into the growing necessity of visual edge intelligence for applications requiring real-time decision-making. Learn about the advancements in object detection accuracy due to Convolutional Neural Networks (CNNs) and Deep Learning. Discover efforts to reduce processing demands of edge-based CNN inference through a hierarchical framework, enabling object detection in high-resolution video frames while maintaining state-of-the-art accuracy. Gain insights into the validation of these techniques on UAV platforms for car and pedestrian detection applications. The talk covers challenges in object detection, optimization techniques, biological approaches, selective tile processing, and brain-inspired computing. Understand the implementation of EdgeNet and its implications for ultra-low power embedded systems.