ENERGY EFFICIENCY OF CNN ARCHITECTURES USING FPGA

This paper explores the importance of optimizing FPGA accelerators in the context of energy efficiency and enhanced performance for processing convolutional neural networks (CNN) in resource-limited environments. A reconfigurable RTL-level accelerator for CNN-based object detection systems is proposed, focusing on hardware and power consumption optimization techniques. Various aspects such as the importance of CNNs in artificial vision, their fundamental structure, and their acceleration on FPGA-SoC devices are presented. Additionally, the benefits of integrating FPGA with SoC and the design requirements to achieve optimal performance and energy efficiency are discussed. This research highlights the significance of innovative approaches to attaining the desired energy efficiency and performance in resource-constrained environments, such as mobile devices, IoT, and electric vehicles