Imperial College London

ProfessorWayneLuk

Faculty of EngineeringDepartment of Computing

Professor of Computer Engineering
 
 
 
//

Contact

 

+44 (0)20 7594 8313w.luk Website

 
 
//

Location

 

434Huxley BuildingSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@inproceedings{Zhao,
author = {Zhao, R and Ng, H-C and Luk, W and Niu, X},
publisher = {IEEE},
title = {Towards efficient convolutional neural network for domain-specific applications on FPGA},
url = {http://hdl.handle.net/10044/1/62190},
}

RIS format (EndNote, RefMan)

TY  - CPAPER
AB - FPGA becomes a popular technology for imple-menting Convolutional Neural Network (CNN) in recent years.Most CNN applications on FPGA are domain-specific, e.g.,detecting objects from specific categories, in which commonly-used CNN models pre-trained on general datasets may not beefficient enough. This paper presents TuRF, an end-to-end CNNacceleration framework to efficiently deploy domain-specific ap-plications on FPGA by transfer learning that adapts pre-trainedmodels to specific domains, replacing standard convolution layerswith efficient convolution blocks, and applying layer fusion toenhance hardware design performance. We evaluate TuRF bydeploying a pre-trained VGG-16 model for a domain-specificimage recognition task onto a Stratix V FPGA. Results showthat designs generated by TuRF achieve better performance thanprior methods for the original VGG-16 and ResNet-50 models,while for the optimised VGG-16 model TuRF designs are moreaccurate and easier to process.
AU - Zhao,R
AU - Ng,H-C
AU - Luk,W
AU - Niu,X
PB - IEEE
TI - Towards efficient convolutional neural network for domain-specific applications on FPGA
UR - http://hdl.handle.net/10044/1/62190
ER -