Efficient CNN Accelerator with FPGA for eye disorders Prediction
Аннотация
As deep learning technologies are used more frequently, it has become possible to obtain effective Convolutional Neural Network (CNN) inference on edge devices with limited resources and turn into a major problem. CNN inference model deployment increasingly relies on Field-Programmable Gate Arrays (FPGA). This research suggests a convolutional operation accelerator based on FPGA that gives the host flexible control over computation tasks via opcodes and configuration parameters..CNN is a artificial neural network is used for feature extraction. CNN accelerator is a innovative hardware used for increasing performance of CNN for classification as well as regression. Proposed CNN accelerator uses Finite impulse response filter(FIR) for preprocessing, Convolution layer for feature extraction and Fully connected layer for prediction.CNNs’ high power consumption and latency when running on traditional hardware architectures like CPUs and GPUs make them less suitable for low-power, high-performance applications. Our study’s CNN accelerator solution is built on an FPGA, which offers low power consumption, flexibility, and high computational throughput.