Date of Publication :5th June 2025

Abstract: This paper presents an FPGA-based implementation of a Convolutional Neural Network (CNN), leveraging the hardware acceleration capabilities of Field-Programmable Gate Arrays (FPGAs) to optimize deep learning computations. With the increasing demand for real-time processing in image recognition and video analysis, FPGA implementations provide an efficient alternative to conventional CPU-based architectures. The proposed modular CNN design enables flexibility, scalability, and improved computational efficiency. Key architectural components such as convolution, pooling, and fully connected layers are structured to maximize parallelism while optimizing resource utilization. Performance evaluations demonstrate the effectiveness of this implementation, showing improvements in processing speed and resource efficiency. The results affirm the suitability of FPGA-based CNNs for high-performance computing applications in deep learning.

Reference :