Image Classification CNNs using the FINN Engine for SRAM-based APSoC in Satellite Applications

This study investigates the performance of two Convolutional Neural Networks (CNNs) designed for aerial image classification. We modify to these CNNs by adjusting parameters such as quantization and parallelism to facilitate their implementation using the FINN engine optimized for the Zynq-7000 APSoC. The goal is to assess the effects of different quantization levels, network topologies, and architectural parallelism on the area, computing performance, and power consumption within the Zynq APSoC. Results illustrate various trade-offs between resource usage, computational performance, and power consumption, all while meeting predefined accuracy requirements.