关键词: Caffe 框架, ZYNQ, 卷积神经网络, 物体分类

Abstract: With the rapid development of deep learning, neural networks and deep learning algorithms have been widely used in image processing.This article explores the FPGA-based neural network acceleration design, and builds an object recognition system based on Convolutional Neural Network of convolutional architecture for fast feature embedding (Caffe)framework,which implements neural network with Zynq-7000 series heterogeneous multicore architecture chip.Transplantation of neural network models and parameters,construction of multi-layer structured neural network, computational intensity analysis and hardware acceleration design are realized.The results show that the Caffe framework object classification system based on heterogeneous multicore platform realizes the function of object recognition and classification, and the recognition speed goes beyond the traditional CPU architecture,which provides a new idea for the subsequent in-depth study.

Key words: Caffe framework, ZYNQ, Convolutional Neural Network, object classification