期刊
APPLIED SCIENCES-BASEL
卷 13, 期 1, 页码 -出版社
MDPI
DOI: 10.3390/app13010093
关键词
convolutional neural networks (CNN); hardware accelerators; systolic array; field programmable gate arrays (FPGA); embedded systems
This paper presents a flexible convolver that can adapt to different convolution layer configurations of state-of-the-art CNNs. The adaptability is achieved by using two proposed programmable components. A Programmable Line Buffer based on Programmable Shift Registers generates the required convolution masks for each processed CNN layer. The convolution layer computing is performed through a proposed programmable systolic array. The experimental results show that the proposed computing method allows for the processing of any CNN without requiring special adaptation for a specific application.
This paper introduces a flexible convolver capable of adapting to the different convolution layer configurations of state-of-the-art Convolution Neural Networks (CNNs). The use of two proposed programmable components achieves this adaptability. A Programmable Line Buffer (PLB) based on Programmable Shift Registers (PSRs) allows the generation of the required convolution masks required for each processed CNN layer. The convolution layer computing is performed through a proposed programmable systolic array configured according to the target device resources. In order to maximize the device resource usage and to achieve a shortened processing time, the filter, data, and loop parallelisms are leveraged. These characteristics allow the described architecture to be scalable and implemented on any FPGA device targeting different applications. The convolver description was written in VHDL using the Intel Cyclone V 5CSXFC6D6F31C6N device as a reference. The experimental results show that the proposed computing method allows the processing of any CNN without requiring special adaptation for a specific application since the standard convolution algorithm is used. The proposed flexible convolver achieves competitive performance compared with those reported in related works.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据