Hardware Implementation of a Convolutional Neural Network

The paper is devoted to the hardware implementation of a convolutional neural network for image recognition systems. The goal of the work is to implement image recognition CNN in FPGA. During the research, algorithms and computer programs for image recognition at the hardware level were developed using the model-driven design approach. The developed neural network consists of 7 layers with a prediction accuracy of 99.54%. The neural network model was optimized by reducing the number of filters in the layers and changing the kernel size, resulting in a decreased prediction accuracy to 86.96%. Logical synthesis of the generated HDL code was performed. The simulation results matched the theoretical assumptions. The obtained results have significant practical value in various scientific fields including medicine, autonomous driving, robotics, quality control, etc.