The ever-growing number of live video feeds in modern establishments has made it necessary to move the processing closer to the imaging sensors. This so called “Edge Computing” enables low latency, privacy, security, and scalability for AI applications in various domains. However, computing at the edge faces many challenges, such as limited energy and computing resources. To this end, a variety of hardware solutions are being considered. At one end of this spectrum is the trivial low resolution object detection on the smallest microcontrollers e.g. FOMO on ESP32-CAM microcontroller. At the other end, NVIDIA Jetson Orin/Xavier embedded GPUs offer high performance in small form factor albeit at a higher cost. RISC-V microprocessors (e.g. Kendryte K210) with dedicated Neural Processing Units (NPUs) offer a middle ground albeit at the expense of lesser portability. This talk will cover a brief overview of these alternative solutions followed by a group discussion on the merits of each approach with respect to cost, learning curve and computational/energy efficiency.